File size: 55,958 Bytes
fd703d8 cc49249 f1a1f51 76bad09 f1a1f51 76bad09 f1a1f51 76bad09 cc49249 fd703d8 f29f15d fd703d8 7914f84 fd703d8 7914f84 fd703d8 f29f15d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 | ---
license: other
license_name: kimi-k3
library_name: transformers
pipeline_tag: image-text-to-text
tags:
- vLLM
- cubic-quantization
- W2A8
- W3A8
- W4A8
- W2A16
- W3A16
- W4A16
- multimodal
base_model:
- moonshotai/Kimi-K3
base_model_relation: quantized
---
# Kimi-K3-Cubic-2.5Bit
Base model: [moonshotai/Kimi-K3](https://huggingface.co/moonshotai/Kimi-K3)
**2.8T parameters Β· 2.5-bit experts Β· one 8 Γ H200 node Β· ~700K context**
<div align="center">
<h3>Keep Open Quantized Models Accessible</h3>
<p>
Large model checkpoints are expensive to host, and Hugging Face storage is becoming the main bottleneck.
</p>
<p>
Your support helps keep quantized models free and easy to download for researchers, developers, students,
and small teams without large GPU budgets.
</p>
<p>
Every contribution goes toward storage costs, helping more people experiment, evaluate, and build with open models.
</p>
<p>
<a href="https://ko-fi.com/quanttrio">
<img src="https://storage.ko-fi.com/cdn/kofi5.png" alt="Support us on Ko-fi" height="54">
</a>
</p>
</div>
Kimi-K3-Cubic-2.5Bit is the first public showcase of **Cubic quantization**,
exploring practical ultra-low-bit, high-concurrency LLM inference.
Cubic uses parameterized, non-uniform levels that adapt to each weight group,
designed to challenge conventional integer and floating-point formats at
ultra-low precision while relaxing the usual need for small group sizes; most
quantized layers in this model use group size 512.
Technical report: [arXiv:2608.06763](https://arxiv.org/abs/2608.06763).
- **57 output tok/s** β single-request short-context decode
- **800β900+ output tok/s** β 64-way concurrency, without dSpark or other
speculative decoding
- **W1βW8 runtime kernels** β native SM80/86/89/90/90a/100/120 coverage
The export is fully data-free: no calibration dataset or activation samples
were used. The complete converter is included as
[`quantize_k3.py`](./quantize_k3.py). The same packed checkpoint runs in either
weight-only A16 or dynamic-A8 mode through the
[QuantTrio vLLM Cubic runtime](https://github.com/QuantTrio/vllm-cubic).
This checkpoint itself uses mixed W2, W3, and W4 expert weights with an audited
effective width of **2.4986 bits**.
### γQuantization Policyγ
| Scope | Format |
|-------|--------|
| Dense layer 0 | Source dtype (BF16) |
| MoE layers 1β3 routed experts | Cubic W3, group size 256 |
| MoE layers 4β32 routed experts | Cubic W3, group size 512 |
| MoE layers 33β91 routed experts | Cubic W2, group size 512 |
| MoE layer 92 routed experts | Cubic W4, group size 512 |
| Attention/KDA/MLA weights | Source dtype (BF16) |
| Shared experts and ordinary MLP weights | Source dtype (BF16) |
| Vision tower, multimodal projector, embeddings, norms, and LM head | Source dtype |
`Loss` is the scale-free NRMSE of groupwise weight reconstruction, not a
downstream task-accuracy score.
| Format | Loss | Clipped values |
|--------|-----:|---------------:|
| W2, group size 512 | 0.451251 | 12.2307% |
| W3, group size 256 | 0.179608 | 2.6971% |
| W3, group size 512 | 0.181007 | 2.9948% |
| W4, group size 512 | 0.043047 | 0.5635% |
### γDependencies / Installationγ
- Runtime: [vLLM Cubic v0.26.1+cubic.20260805](https://github.com/QuantTrio/vllm-cubic/releases/tag/v0.26.1%2Bcubic.20260805)
- Python: 3.10β3.14
- Platform: Linux x86-64, PyTorch 2.13.0, CUDA 13.0
- Verified: Python 3.12, 8 Γ H200, TP8 + EP
Install:
```bash
uv venv --python 3.12 .venv
source .venv/bin/activate
uv pip install \
'https://github.com/QuantTrio/vllm-cubic/releases/download/v0.26.1%2Bcubic.20260805/vllm-0.26.1%2Bcubic.20260805-cp38-abi3-linux_x86_64.whl'
```
### γQuantization Scriptγ
- Script: [`quantize_k3.py`](./quantize_k3.py) ([source](https://github.com/QuantTrio/vllm-cubic/blob/main/examples/quantization/quantize_k3.py))
- Input: original `moonshotai/Kimi-K3` MXFP4 checkpoint
- Method: data-free; dynamic-A8 carrier correction on by default (`--disable-a8-correction` to disable)
- Time: approximately **35β40 minutes on 8 Γ H200** with fast local storage
- Dependencies: included with vLLM Cubic; standalone: `pip install torch safetensors regex`
- Output: destination path must not already exist
```bash
python -u quantize_k3.py \
--source __YOUR_PATH__/moonshotai/Kimi-K3 \
--output __YOUR_PATH__/Kimi-K3-Cubic-2.5Bit \
--devices cuda:0,cuda:1,cuda:2,cuda:3,cuda:4,cuda:5,cuda:6,cuda:7
```
### γvLLM Startup Commandγ
The following command enables dynamic A8 Cubic execution. Change
`VLLM_CUBIC_DYNAMIC_A8` to `0` to run the same checkpoint in weight-only A16
mode.
```bash
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
export VLLM_CUBIC_DYNAMIC_A8=1
vllm serve \
__YOUR_PATH__/Kimi-K3-Cubic-2.5Bit \
--served-model-name Kimi-K3-Cubic-2.5Bit \
--trust-remote-code \
--quantization cubic \
--kv-cache-dtype fp8_q16 \
--gpu-memory-utilization 0.985 \
--tensor-parallel-size 8 \
--enable-expert-parallel \
--mm-encoder-tp-mode data \
--max-model-len auto \
--max-num-seqs 128 \
--max-num-batched-tokens 2048 \
--enable-prefix-caching \
--enable-auto-tool-choice \
--tool-call-parser kimi_k3 \
--reasoning-parser kimi_k3 \
--host 0.0.0.0 \
--port 8000 \
--disable-uvicorn-access-log
```
Observed performance for this command on 8 Γ NVIDIA H200 141GB and 8 Γ NVIDIA H20 141GB
(2026-08-05) is:
### γdocker with vLLM Startup Commandγ
```bash
docker pull swr.cn-north-4.myhuaweicloud.com/desaysv/vllm-cubic:0.26.1-cubic.20260805
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
export VLLM_CUBIC_DYNAMIC_A8=1
export FLASHINFER_DISABLE_VERSION_CHECK=1
vllm serve \
__YOUR_PATH__/Kimi-K3-Cubic-2.5Bit \
--served-model-name Kimi-K3-Cubic-2.5Bit \
--trust-remote-code \
--quantization cubic \
--kv-cache-dtype fp8_q16 \
--gpu-memory-utilization 0.985 \
--tensor-parallel-size 8 \
--enable-expert-parallel \
--mm-encoder-tp-mode data \
--max-model-len auto \
--max-num-seqs 128 \
--max-num-batched-tokens 2048 \
--enable-prefix-caching \
--enable-auto-tool-choice \
--tool-call-parser kimi_k3 \
--reasoning-parser kimi_k3 \
--host 0.0.0.0 \
--port 8000 \
--disable-uvicorn-access-log
```
*skip deepgemm error ,its not effect start
### γbenchmarkγ10*8K/1K
| Item | Observed value |
|------|---------------:|
| Total requests | 10 |
| Successful requests | 10 |
| Failed requests | 0 |
| Unfinished requests | 0 |
| Average concurrency | 10 |
| Total duration | 209.75 s |
| Total throughput | 464.11 tokens/s |
| Input throughput | 684.34 tokens/s |
| Output throughput | 51.07 tokens/s |
| Average request latency | 204.23 s |
| Average TTFT | 69,607.26 ms |
| Average TPOT | 164.76 ms |
| Average ITL | 95.25 ms |
| TTFT (P50) | 63,297.56 ms |
| TTFT (P90) | 128,423.96 ms |
| TTFT (P99) | 128,423.96 ms |
| ITL (P50) | 81.85 ms |
| ITL (P90) | 178.56 ms |
| ITL (P99) | 178.56 ms |
| TPOT (P50) | 154.40 ms |
| TPOT (P90) | 206.35 ms |
| TPOT (P99) | 206.35 ms |
| Latency (P50) | 204.10 s |
| Latency (P90) | 208.45 s |
| Latency (P99) | 208.45 s |
| Input tokens per request | 8,088 |
| Output tokens per request | 1,000 |
### γbenchmarkγ10*64K/3000
| Item | Observed value |
|------|---------------:|
| Total requests | 10 |
| Successful requests | 10 |
| Failed requests | 0 |
| Unfinished requests | 0 |
| Average concurrency | 9 |
| Total duration | 1,461.29 s |
| Total throughput | 519.15 tokens/s |
| Input throughput | 513.63 tokens/s |
| Output throughput | 23.21 tokens/s |
| Average request latency | 1,387.96 s |
| Average TTFT | 882,028.37 ms |
| Average TPOT | 334.97 ms |
| Average ITL | 107.57 ms |
| TTFT (P50) | 604,938.80 ms |
| TTFT (P90) | 1,352,678.48 ms |
| TTFT (P99) | 1,352,678.48 ms |
| ITL (P50) | 61.83 ms |
| ITL (P90) | 313.74 ms |
| ITL (P99) | 313.74 ms |
| TPOT (P50) | 378.19 ms |
| TPOT (P90) | 465.73 ms |
| TPOT (P99) | 465.73 ms |
| Latency (P50) | 1,394.02 s |
| Latency (P90) | 1,460.10 s |
| Latency (P99) | 1,460.10 s |
| Input tokens per request | 64,088 |
| Output tokens per request | 3,000 |
| Item | Observed value |
|------|---------------:|
| Dynamic Cubic mode | A8 |
| Parallel topology | TP8 + EP |
| KV-cache dtype | `fp8_q16` |
| Effective maximum model length | 728,640 tokens |
| Single-request short-context decode | approximately 57 output tok/s |
| 64-request aggregate decode | approximately 800β900+ output tok/s |
| Speculative decoding | None; dSpark was not used |
The throughput values are stable decode-window measurements from short-prompt,
long-output text requests, rather than end-to-end request throughput. Prompt
length, generated context length, sampling, concurrent prefill, GPU clocks, and
other workloads can change the result. Because `--max-model-len auto` derives
the limit from free memory measured during startup, the effective token count
can also vary slightly between launches.
`fp8_q16` is an optional query-preserving FP8 KV-cache mode from the Cubic
runtime. It keeps attention queries in the model dtype and automatically
selects a compatible attention backend. To use a conventional BF16 KV cache,
replace it with `--kv-cache-dtype auto`; this reduces the available context
length but provides the highest-precision KV storage.
The first launch may spend several minutes compiling Triton kernels,
calibrating Cubic kernel choices for the tensors and GPU type in use, and
capturing CUDA graphs. Calibration results are persisted in the Triton cache
and reused by later launches on compatible devices.
### γValidation Statusγ
- 8 Γ H200 TP8 + EP model loading and text generation verified.
- Weight-only A16 and dynamic-A8 execution verified with the same checkpoint.
- `fp8_q16` and BF16 KV-cache serving paths verified.
- No formal downstream benchmark is claimed for this quantized checkpoint;
evaluate quality and performance for your own workloads before deployment.
### γModel Filesγ
| Item | Value |
|------|-------|
| Model data | 899 GiB |
| Safetensors shards | 341 |
| Last updated | 2026-08-05 |
### γModel Downloadγ
```python
from huggingface_hub import snapshot_download
snapshot_download(
"QuantTrio/Kimi-K3-Cubic-2.5Bit",
local_dir="your_local_path/Kimi-K3-Cubic-2.5Bit",
)
```
### γLicenseγ
This quantized checkpoint is derived from Kimi K3 and remains subject to the
[Kimi K3 License](https://huggingface.co/moonshotai/Kimi-K3/blob/main/LICENSE).
The QuantTrio vLLM Cubic runtime is distributed separately under its own
repository license.
### γBase Model Overviewγ
The following material is reproduced from the upstream Kimi K3 model card and
describes the base model. Upstream MXFP4 deployment instructions do not apply
to this Cubic checkpoint; use the Cubic installation and launch command above.
<div align="center">
<picture>
<img src="assets/kimi-logo.png" width="30%" alt="Kimi K3">
</picture>
</div>
<hr>
<div align="center" style="line-height:1">
<a href="https://www.kimi.com" target="_blank"><img alt="Chat" src="https://img.shields.io/badge/π€%20Chat-Kimi%20K3-ff6b6b?color=1783ff&logoColor=white"/></a>
<a href="https://www.moonshot.ai" target="_blank"><img alt="Homepage" src="https://img.shields.io/badge/Homepage-Moonshot%20AI-white?logo=Kimi&logoColor=white"/></a>
</div>
<div align="center" style="line-height: 1;">
<a href="https://huggingface.co/moonshotai" target="_blank"><img alt="Hugging Face" src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Moonshot%20AI-ffc107?color=ffc107&logoColor=white"/></a>
<a href="https://twitter.com/kimi_moonshot" target="_blank"><img alt="Twitter Follow" src="https://img.shields.io/badge/Twitter-Kimi.ai-white?logo=x&logoColor=white"/></a>
<a href="https://discord.gg/TYU2fdJykW" target="_blank"><img alt="Discord" src="https://img.shields.io/badge/Discord-Kimi.ai-white?logo=discord&logoColor=white"/></a>
<a href="https://modelscope.cn/organization/moonshotai" target="_blank"><img alt="ModelScope" src="https://img.shields.io/badge/ModelScope-Moonshot%20AI-white?labelColor=rgb(99%2C%2074%2C%255)"/></a>
</div>
<div align="center" style="line-height: 1;">
<a href="https://huggingface.co/moonshotai/Kimi-K3/blob/main/LICENSE"><img alt="License" src="https://img.shields.io/badge/License-Kimi_K3-f5de53?&color=f5de53"/></a>
</div>
<p align="center">
π° <a href="https://www.kimi.com/blog/kimi-k3">Tech Blog</a> | <b>π <a href="https://github.com/MoonshotAI/Kimi-K3/blob/main/k3_tech_report.pdf">Full Report</a></b>
</p>
## 1. Model Introduction
Kimi K3 is an open-weight, native multimodal agentic model and our most capable model to date. It is a 2.8T-parameter model built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), with native vision capabilities and a 1-million-token context window. It is the world's first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning.
### Key Features
- **New Architecture**: Kimi K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), and scales up MoE sparsity with a Stable LatentMoE framework that activates 16 out of 896 experts β yielding an approximate 2.5Γ improvement in overall scaling efficiency over Kimi K2.
- **Long-Horizon Coding**: Operating with minimal human oversight, Kimi K3 sustains long engineering sessions, navigates massive repositories, and orchestrates terminal tools β from GPU kernel optimization and compiler development to vision-in-the-loop game dev, CAD, and even chip design.
- **Agentic Knowledge Work**: Kimi K3 advances end-to-end knowledge work, producing deep research with interactive visualizations, widgets and dashboards, and motion design and video editing, powered by its native multimodal architecture.
- **Native Multimodality & Long Context**: Kimi K3 understands text, images, and video within the same model, and supports a 1-million-token context window.
- **Open Frontier Weights**: We release the full Kimi K3 model weights under the Kimi K3 License, making frontier intelligence openly available for research, deployment, and further innovation.
## 2. Model Summary
<div align="center">
<table>
<tbody>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Architecture</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">Mixture-of-Experts (MoE)</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Total Parameters</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">2.8T</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Activated Parameters</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">104B</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Number of Layers</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">93</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Number of Dense Layers</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">1</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Attention-Layer Composition</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">69 KDA + 24 Gated MLA</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Attention Hidden Dimension</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">7168</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Number of Attention Heads</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">96</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Latent MoE Dimension</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">3584</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>MoE Hidden Dimension</strong> (per Expert)</td>
<td align="center" style="vertical-align: middle; text-align: center">3072</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Number of Experts</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">896</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Selected Experts per Token</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">16</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Number of Shared Experts</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">2</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Vocabulary Size</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">160K</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Context Length</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">1048576</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Attention Mechanism</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">KDA & Gated MLA</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Activation Function</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">SiTU-GLU</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Vision Encoder</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">MoonViT-V2</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Parameters of Vision Encoder</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">401M</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Quantization</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">MXFP4 weights / MXFP8 activations<br>(quantization-aware training)</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center"><strong>Modality</strong></td>
<td align="center" style="vertical-align: middle; text-align: center">Text, Image</td>
</tr>
</tbody>
</table>
</div>
## 3. Evaluation Results
<div align="center">
<table>
<thead>
<tr>
<th align="center" style="text-align: center">Benchmark</th>
<th align="center" style="text-align: center"><sup>Kimi K3<br><sup>(max)</sup></sup></th>
<th align="center" style="text-align: center"><sup>Claude Fable 5<br><sup>(max, w/ fallback)</sup></sup></th>
<th align="center" style="text-align: center"><sup>GPT-5.6 Sol<br><sup>(max)</sup></sup></th>
<th align="center" style="text-align: center"><sup>Claude Opus 4.8<br><sup>(max)</sup></sup></th>
<th align="center" style="text-align: center"><sup>GPT-5.5<br><sup>(xhigh)</sup></sup></th>
<th align="center" style="text-align: center"><sup>GLM-5.2<br><sup>(max)</sup></sup></th>
</tr>
</thead>
<tbody>
<tr>
<td align="center" colspan=7 style="text-align: center"><strong>Reasoning & Knowledge</strong></td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">GPQA Diamond</td>
<td align="center" style="vertical-align: middle; text-align: center">93.5</td>
<td align="center" style="vertical-align: middle; text-align: center">92.6</td>
<td align="center" style="vertical-align: middle; text-align: center">94.1</td>
<td align="center" style="vertical-align: middle; text-align: center">91.0</td>
<td align="center" style="vertical-align: middle; text-align: center">93.5</td>
<td align="center" style="vertical-align: middle; text-align: center">91.2</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">CritPt</td>
<td align="center" style="vertical-align: middle; text-align: center">23.4</td>
<td align="center" style="vertical-align: middle; text-align: center">28.6</td>
<td align="center" style="vertical-align: middle; text-align: center">32.3</td>
<td align="center" style="vertical-align: middle; text-align: center">20.9</td>
<td align="center" style="vertical-align: middle; text-align: center">27.1</td>
<td align="center" style="vertical-align: middle; text-align: center">20.9</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">AA-LCR</td>
<td align="center" style="vertical-align: middle; text-align: center">74.7</td>
<td align="center" style="vertical-align: middle; text-align: center">70.0</td>
<td align="center" style="vertical-align: middle; text-align: center">73.7</td>
<td align="center" style="vertical-align: middle; text-align: center">67.7</td>
<td align="center" style="vertical-align: middle; text-align: center">74.3</td>
<td align="center" style="vertical-align: middle; text-align: center">71.3</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">HLE-Full</td>
<td align="center" style="vertical-align: middle; text-align: center">43.5 / 56.0</td>
<td align="center" style="vertical-align: middle; text-align: center">53.3 / 63.0</td>
<td align="center" style="vertical-align: middle; text-align: center">44.5 / 58.0</td>
<td align="center" style="vertical-align: middle; text-align: center">49.8 / 57.9</td>
<td align="center" style="vertical-align: middle; text-align: center">41.4 / 52.2</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" colspan=7 style="text-align: center"><strong>Coding</strong></td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">DeepSWE</td>
<td align="center" style="vertical-align: middle; text-align: center">67.5</td>
<td align="center" style="vertical-align: middle; text-align: center">70.0</td>
<td align="center" style="vertical-align: middle; text-align: center">73.0</td>
<td align="center" style="vertical-align: middle; text-align: center">59.0</td>
<td align="center" style="vertical-align: middle; text-align: center">67.0</td>
<td align="center" style="vertical-align: middle; text-align: center">46.2</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">ProgramBench</td>
<td align="center" style="vertical-align: middle; text-align: center">77.8</td>
<td align="center" style="vertical-align: middle; text-align: center">76.8</td>
<td align="center" style="vertical-align: middle; text-align: center">77.6</td>
<td align="center" style="vertical-align: middle; text-align: center">71.9</td>
<td align="center" style="vertical-align: middle; text-align: center">70.8</td>
<td align="center" style="vertical-align: middle; text-align: center">63.7</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">Terminal-Bench 2.1</td>
<td align="center" style="vertical-align: middle; text-align: center">88.3</td>
<td align="center" style="vertical-align: middle; text-align: center">88.0</td>
<td align="center" style="vertical-align: middle; text-align: center">88.8</td>
<td align="center" style="vertical-align: middle; text-align: center">84.6</td>
<td align="center" style="vertical-align: middle; text-align: center">83.4</td>
<td align="center" style="vertical-align: middle; text-align: center">82.7</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">FrontierSWE</td>
<td align="center" style="vertical-align: middle; text-align: center">81.2</td>
<td align="center" style="vertical-align: middle; text-align: center">86.6</td>
<td align="center" style="vertical-align: middle; text-align: center">71.3</td>
<td align="center" style="vertical-align: middle; text-align: center">66.7</td>
<td align="center" style="vertical-align: middle; text-align: center">64.9</td>
<td align="center" style="vertical-align: middle; text-align: center">67.3</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">SWE-Marathon</td>
<td align="center" style="vertical-align: middle; text-align: center">42.0</td>
<td align="center" style="vertical-align: middle; text-align: center">35.0</td>
<td align="center" style="vertical-align: middle; text-align: center">39.0</td>
<td align="center" style="vertical-align: middle; text-align: center">40.0</td>
<td align="center" style="vertical-align: middle; text-align: center">14.0</td>
<td align="center" style="vertical-align: middle; text-align: center">13.0</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">PostTrainBench</td>
<td align="center" style="vertical-align: middle; text-align: center">36.6</td>
<td align="center" style="vertical-align: middle; text-align: center">41.4</td>
<td align="center" style="vertical-align: middle; text-align: center">34.6</td>
<td align="center" style="vertical-align: middle; text-align: center">34.1</td>
<td align="center" style="vertical-align: middle; text-align: center">28.4</td>
<td align="center" style="vertical-align: middle; text-align: center">34.3</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">MLS-Bench-Lite</td>
<td align="center" style="vertical-align: middle; text-align: center">48.3</td>
<td align="center" style="vertical-align: middle; text-align: center">49.9</td>
<td align="center" style="vertical-align: middle; text-align: center">46.2</td>
<td align="center" style="vertical-align: middle; text-align: center">42.8</td>
<td align="center" style="vertical-align: middle; text-align: center">35.5</td>
<td align="center" style="vertical-align: middle; text-align: center">40.4</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">SciCode</td>
<td align="center" style="vertical-align: middle; text-align: center">58.7</td>
<td align="center" style="vertical-align: middle; text-align: center">60.2</td>
<td align="center" style="vertical-align: middle; text-align: center">56.1</td>
<td align="center" style="vertical-align: middle; text-align: center">53.5</td>
<td align="center" style="vertical-align: middle; text-align: center">56.1</td>
<td align="center" style="vertical-align: middle; text-align: center">50.5</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">Kimi Code Bench 2.0</td>
<td align="center" style="vertical-align: middle; text-align: center">72.9</td>
<td align="center" style="vertical-align: middle; text-align: center">76.9</td>
<td align="center" style="vertical-align: middle; text-align: center">64.8</td>
<td align="center" style="vertical-align: middle; text-align: center">71.7</td>
<td align="center" style="vertical-align: middle; text-align: center">69.0</td>
<td align="center" style="vertical-align: middle; text-align: center">64.2</td>
</tr>
<tr>
<td align="center" colspan=7 style="text-align: center"><strong>Agentic</strong></td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">BrowseComp</td>
<td align="center" style="vertical-align: middle; text-align: center">91.2</td>
<td align="center" style="vertical-align: middle; text-align: center">88.0</td>
<td align="center" style="vertical-align: middle; text-align: center">90.4</td>
<td align="center" style="vertical-align: middle; text-align: center">84.3</td>
<td align="center" style="vertical-align: middle; text-align: center">84.4</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">DeepSearchQA (F1)</td>
<td align="center" style="vertical-align: middle; text-align: center">95.0</td>
<td align="center" style="vertical-align: middle; text-align: center">94.2</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
<td align="center" style="vertical-align: middle; text-align: center">93.1</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">ResearchRubrics</td>
<td align="center" style="vertical-align: middle; text-align: center">76.2</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
<td align="center" style="vertical-align: middle; text-align: center">73.8</td>
<td align="center" style="vertical-align: middle; text-align: center">73.5</td>
<td align="center" style="vertical-align: middle; text-align: center">64.0</td>
<td align="center" style="vertical-align: middle; text-align: center">71.1</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">GDPval-AA v2 (Elo)</td>
<td align="center" style="vertical-align: middle; text-align: center">1686</td>
<td align="center" style="vertical-align: middle; text-align: center">1747</td>
<td align="center" style="vertical-align: middle; text-align: center">1736</td>
<td align="center" style="vertical-align: middle; text-align: center">1593</td>
<td align="center" style="vertical-align: middle; text-align: center">1491</td>
<td align="center" style="vertical-align: middle; text-align: center">1510</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">Toolathlon-Verified</td>
<td align="center" style="vertical-align: middle; text-align: center">76.5</td>
<td align="center" style="vertical-align: middle; text-align: center">77.9</td>
<td align="center" style="vertical-align: middle; text-align: center">74.9</td>
<td align="center" style="vertical-align: middle; text-align: center">76.2</td>
<td align="center" style="vertical-align: middle; text-align: center">73.5</td>
<td align="center" style="vertical-align: middle; text-align: center">59.9</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">MCPMark-Verified</td>
<td align="center" style="vertical-align: middle; text-align: center">94.5</td>
<td align="center" style="vertical-align: middle; text-align: center">87.4</td>
<td align="center" style="vertical-align: middle; text-align: center">92.9</td>
<td align="center" style="vertical-align: middle; text-align: center">76.4</td>
<td align="center" style="vertical-align: middle; text-align: center">92.9</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">MCP-Atlas</td>
<td align="center" style="vertical-align: middle; text-align: center">84.2</td>
<td align="center" style="vertical-align: middle; text-align: center">84.7</td>
<td align="center" style="vertical-align: middle; text-align: center">83.6</td>
<td align="center" style="vertical-align: middle; text-align: center">83.6</td>
<td align="center" style="vertical-align: middle; text-align: center">82.8</td>
<td align="center" style="vertical-align: middle; text-align: center">82.6</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">AutomationBench</td>
<td align="center" style="vertical-align: middle; text-align: center">30.8</td>
<td align="center" style="vertical-align: middle; text-align: center">29.1</td>
<td align="center" style="vertical-align: middle; text-align: center">29.7</td>
<td align="center" style="vertical-align: middle; text-align: center">27.2</td>
<td align="center" style="vertical-align: middle; text-align: center">22.7</td>
<td align="center" style="vertical-align: middle; text-align: center">12.9</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">JobBench</td>
<td align="center" style="vertical-align: middle; text-align: center">54.3</td>
<td align="center" style="vertical-align: middle; text-align: center">57.4</td>
<td align="center" style="vertical-align: middle; text-align: center">45.4</td>
<td align="center" style="vertical-align: middle; text-align: center">48.4</td>
<td align="center" style="vertical-align: middle; text-align: center">38.3</td>
<td align="center" style="vertical-align: middle; text-align: center">43.4</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">AA-Briefcase (Elo)</td>
<td align="center" style="vertical-align: middle; text-align: center">1548</td>
<td align="center" style="vertical-align: middle; text-align: center">1583</td>
<td align="center" style="vertical-align: middle; text-align: center">1495</td>
<td align="center" style="vertical-align: middle; text-align: center">1354</td>
<td align="center" style="vertical-align: middle; text-align: center">1158</td>
<td align="center" style="vertical-align: middle; text-align: center">1260</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">Agents' Last Exam</td>
<td align="center" style="vertical-align: middle; text-align: center">28.3</td>
<td align="center" style="vertical-align: middle; text-align: center">25.7<sup>β </sup></td>
<td align="center" style="vertical-align: middle; text-align: center">29.6</td>
<td align="center" style="vertical-align: middle; text-align: center">27.0</td>
<td align="center" style="vertical-align: middle; text-align: center">26.6</td>
<td align="center" style="vertical-align: middle; text-align: center">20.4</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">APEX-Agents</td>
<td align="center" style="vertical-align: middle; text-align: center">41.0</td>
<td align="center" style="vertical-align: middle; text-align: center">43.3</td>
<td align="center" style="vertical-align: middle; text-align: center">39.9</td>
<td align="center" style="vertical-align: middle; text-align: center">39.4</td>
<td align="center" style="vertical-align: middle; text-align: center">38.5</td>
<td align="center" style="vertical-align: middle; text-align: center">35.6</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">OfficeQA Pro</td>
<td align="center" style="vertical-align: middle; text-align: center">63.3</td>
<td align="center" style="vertical-align: middle; text-align: center">69.9</td>
<td align="center" style="vertical-align: middle; text-align: center">63.2</td>
<td align="center" style="vertical-align: middle; text-align: center">63.9</td>
<td align="center" style="vertical-align: middle; text-align: center">60.9</td>
<td align="center" style="vertical-align: middle; text-align: center">41.4</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">SpreadsheetBench 2</td>
<td align="center" style="vertical-align: middle; text-align: center">34.8</td>
<td align="center" style="vertical-align: middle; text-align: center">34.7</td>
<td align="center" style="vertical-align: middle; text-align: center">32.4</td>
<td align="center" style="vertical-align: middle; text-align: center">31.6</td>
<td align="center" style="vertical-align: middle; text-align: center">29.1</td>
<td align="center" style="vertical-align: middle; text-align: center">28.1</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">OSWorld-Verified</td>
<td align="center" style="vertical-align: middle; text-align: center">84.8</td>
<td align="center" style="vertical-align: middle; text-align: center">85.0</td>
<td align="center" style="vertical-align: middle; text-align: center">83.0</td>
<td align="center" style="vertical-align: middle; text-align: center">83.4</td>
<td align="center" style="vertical-align: middle; text-align: center">79.0</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">OSWorld 2.0</td>
<td align="center" style="vertical-align: middle; text-align: center">58.3</td>
<td align="center" style="vertical-align: middle; text-align: center">66.1</td>
<td align="center" style="vertical-align: middle; text-align: center">62.6</td>
<td align="center" style="vertical-align: middle; text-align: center">55.7</td>
<td align="center" style="vertical-align: middle; text-align: center">49.5</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">SaaS-Bench</td>
<td align="center" style="vertical-align: middle; text-align: center">60.1</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
<td align="center" style="vertical-align: middle; text-align: center">61.4</td>
<td align="center" style="vertical-align: middle; text-align: center">56.1</td>
<td align="center" style="vertical-align: middle; text-align: center">43.8</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">ΟΒ³-Banking</td>
<td align="center" style="vertical-align: middle; text-align: center">33.4</td>
<td align="center" style="vertical-align: middle; text-align: center">26.8</td>
<td align="center" style="vertical-align: middle; text-align: center">33.0</td>
<td align="center" style="vertical-align: middle; text-align: center">27.6</td>
<td align="center" style="vertical-align: middle; text-align: center">31.3</td>
<td align="center" style="vertical-align: middle; text-align: center">26.8</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">Harvey Lab-AA</td>
<td align="center" style="vertical-align: middle; text-align: center">94.6</td>
<td align="center" style="vertical-align: middle; text-align: center">93.6</td>
<td align="center" style="vertical-align: middle; text-align: center">87.2</td>
<td align="center" style="vertical-align: middle; text-align: center">91.1</td>
<td align="center" style="vertical-align: middle; text-align: center">86.3</td>
<td align="center" style="vertical-align: middle; text-align: center">91.0</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">CorpFin v2</td>
<td align="center" style="vertical-align: middle; text-align: center">71.6</td>
<td align="center" style="vertical-align: middle; text-align: center">71.8</td>
<td align="center" style="vertical-align: middle; text-align: center">64.4</td>
<td align="center" style="vertical-align: middle; text-align: center">66.7</td>
<td align="center" style="vertical-align: middle; text-align: center">68.4</td>
<td align="center" style="vertical-align: middle; text-align: center">66.1</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">Finance Agent v2</td>
<td align="center" style="vertical-align: middle; text-align: center">54.4</td>
<td align="center" style="vertical-align: middle; text-align: center">56.3</td>
<td align="center" style="vertical-align: middle; text-align: center">53.8</td>
<td align="center" style="vertical-align: middle; text-align: center">53.9</td>
<td align="center" style="vertical-align: middle; text-align: center">51.8</td>
<td align="center" style="vertical-align: middle; text-align: center">49.7</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">Legal Research Bench</td>
<td align="center" style="vertical-align: middle; text-align: center">44.2</td>
<td align="center" style="vertical-align: middle; text-align: center">49.5</td>
<td align="center" style="vertical-align: middle; text-align: center">48.1</td>
<td align="center" style="vertical-align: middle; text-align: center">43.8</td>
<td align="center" style="vertical-align: middle; text-align: center">40.4</td>
<td align="center" style="vertical-align: middle; text-align: center">31.3</td>
</tr>
<tr>
<td align="center" colspan=7 style="text-align: center"><strong>Vision</strong></td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">WorldVQA ForceAnswer</td>
<td align="center" style="vertical-align: middle; text-align: center">51.0</td>
<td align="center" style="vertical-align: middle; text-align: center">56.7</td>
<td align="center" style="vertical-align: middle; text-align: center">41.8</td>
<td align="center" style="vertical-align: middle; text-align: center">39.1</td>
<td align="center" style="vertical-align: middle; text-align: center">38.5</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">OmniDocBench</td>
<td align="center" style="vertical-align: middle; text-align: center">91.1</td>
<td align="center" style="vertical-align: middle; text-align: center">89.8</td>
<td align="center" style="vertical-align: middle; text-align: center">85.8</td>
<td align="center" style="vertical-align: middle; text-align: center">87.9</td>
<td align="center" style="vertical-align: middle; text-align: center">89.4</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">PerceptionBench</td>
<td align="center" style="vertical-align: middle; text-align: center">58.5</td>
<td align="center" style="vertical-align: middle; text-align: center">57.2</td>
<td align="center" style="vertical-align: middle; text-align: center">59.7</td>
<td align="center" style="vertical-align: middle; text-align: center">47.2</td>
<td align="center" style="vertical-align: middle; text-align: center">55.8</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">Video-MME (w. sub)</td>
<td align="center" style="vertical-align: middle; text-align: center">90.0</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
<td align="center" style="vertical-align: middle; text-align: center">89.5</td>
<td align="center" style="vertical-align: middle; text-align: center">86.0</td>
<td align="center" style="vertical-align: middle; text-align: center">89.3</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">MMVU</td>
<td align="center" style="vertical-align: middle; text-align: center">82.1</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
<td align="center" style="vertical-align: middle; text-align: center">81.2</td>
<td align="center" style="vertical-align: middle; text-align: center">79.2</td>
<td align="center" style="vertical-align: middle; text-align: center">81.7</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">BabyVision w/ python</td>
<td align="center" style="vertical-align: middle; text-align: center">85.7</td>
<td align="center" style="vertical-align: middle; text-align: center">90.5</td>
<td align="center" style="vertical-align: middle; text-align: center">88.9</td>
<td align="center" style="vertical-align: middle; text-align: center">81.2</td>
<td align="center" style="vertical-align: middle; text-align: center">83.6</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">MMMU-Pro</td>
<td align="center" style="vertical-align: middle; text-align: center">81.6 / 83.4</td>
<td align="center" style="vertical-align: middle; text-align: center">81.2 / 86.5</td>
<td align="center" style="vertical-align: middle; text-align: center">83.0 / 84.6</td>
<td align="center" style="vertical-align: middle; text-align: center">78.9 / 82.7</td>
<td align="center" style="vertical-align: middle; text-align: center">81.2 / 83.2</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">CharXiv (RQ)</td>
<td align="center" style="vertical-align: middle; text-align: center">84.8 / 91.3</td>
<td align="center" style="vertical-align: middle; text-align: center">88.9 / 93.5</td>
<td align="center" style="vertical-align: middle; text-align: center">84.6 / 89.1</td>
<td align="center" style="vertical-align: middle; text-align: center">80.5 / 89.9</td>
<td align="center" style="vertical-align: middle; text-align: center">84.1 / 89.0</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">MathVision</td>
<td align="center" style="vertical-align: middle; text-align: center">94.3 / 97.8</td>
<td align="center" style="vertical-align: middle; text-align: center">94.8 / 98.6</td>
<td align="center" style="vertical-align: middle; text-align: center">95.8 / 97.8</td>
<td align="center" style="vertical-align: middle; text-align: center">86.7 / 97.1</td>
<td align="center" style="vertical-align: middle; text-align: center">92.2 / 96.8</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
<tr>
<td align="center" style="vertical-align: middle; text-align: center">ZeroBench (pass@5)</td>
<td align="center" style="vertical-align: middle; text-align: center">23.0 / 41.0</td>
<td align="center" style="vertical-align: middle; text-align: center">23.0 / 46.0</td>
<td align="center" style="vertical-align: middle; text-align: center">17.0 / 35.0</td>
<td align="center" style="vertical-align: middle; text-align: center">17.0 / 34.0</td>
<td align="center" style="vertical-align: middle; text-align: center">22.0 / 41.0</td>
<td align="center" style="vertical-align: middle; text-align: center">β</td>
</tr>
</tbody>
</table>
</div>
<details>
<summary><b>Footnotes</b></summary>
All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. For single-step tasks, such as GPQA Diamond, HLE-Full, and vision benchmarks without tools, we set top-p = 0.95; for agentic tasks, we set top-p = 1.0. For HLE-Full, MMMU-Pro, CharXiv (RQ), MathVision, and ZeroBench, each cell reports the scores without and with tool augmentation (general tools for HLE-Full, Python for the vision benchmarks), in that order.
1. **Reasoning & knowledge benchmarks**
- **CritPt and AA-LCR.** Scores are cited from [Artificial Analysis](https://artificialanalysis.ai/) as of July 23, 2026.
2. **Coding benchmarks**
- **DeepSWE.** Kimi K3 is evaluated with the Kimi Code harness. The GLM-5.2 score is taken from the [GLM-5.2 release blog](https://z.ai/blog/glm-5.2); all remaining scores are from the official [DeepSWE leaderboard](https://deepswe.datacurve.ai/), under which Kimi K3 attains 67.3 with the mini-SWE-agent harness. We report the DeepSWE v1.1 tasks.
- **Terminal-Bench 2.1.** Kimi K3 is evaluated with the Kimi Code harness. For all other models, we report the best score across harnesses: GLM-5.2 with Claude Code ([GLM-5.2 release blog](https://z.ai/blog/glm-5.2)); Claude Opus 4.8 and Claude Fable 5 with Terminus 2 ([Artificial Analysis](https://artificialanalysis.ai/evaluations/terminalbench-v2-1)); GPT-5.5 and GPT-5.6 Sol with Codex ([OpenAI](https://openai.com/index/previewing-gpt-5-6-sol/)).
- **ProgramBench.** Kimi K3 is evaluated with the Kimi Code harness. The GLM-5.2 score is from the [GLM-5.2 release blog](https://z.ai/blog/glm-5.2); all other scores are from [Vals AI](https://www.vals.ai/benchmarks/programbench).
- **SWE-Marathon.** Kimi K3, Claude Opus 4.8, and Claude Fable 5 are evaluated with the Claude Code harness; GPT-5.6 Sol is evaluated with the Codex harness. The GLM-5.2 score is from the [GLM-5.2 release blog](https://z.ai/blog/glm-5.2). Our evaluation is based on an H20-calibrated branch of the [official tasks](https://www.swe-marathon.org/) as of July 9, 2026, prior to the final v1.1 release: the Docker images, performance gates, and reference oracles for the GPU tasks have been recalibrated for H20, while the correctness and anti-cheat validators remain unchanged. Additionally, Claude Fable 5 hit fallbacks on 35% of the tasks in our evaluation, which may have negatively impacted its measured performance.
- **FrontierSWE.** Kimi K3 is evaluated with the Kimi Code harness and GPT-5.6 Sol with the Codex harness; all other results are from [FrontierSWE](https://www.frontierswe.com/). Dominance scores are recomputed from the raw scores using the official evaluation script and are current as of July 16, 2026.
- **PostTrainBench.** Scores for GLM-5.2, GPT-5.5, and Claude Opus 4.8 are adopted from the official [PostTrainBench](https://posttrainbench.com/) results. Kimi K3, Claude Fable 5, and GPT-5.6 Sol are evaluated with the official Harbor implementation at maximum reasoning effort, averaged over three runs on H20 GPUs (instead of H100 in the official setting) β Kimi K3 and Claude Fable 5 with the Claude Code harness, and GPT-5.6 Sol with the Codex harness.
- **MLS-Bench-Lite.** Kimi K3 is evaluated with the Kimi Code harness; GLM-5.2 and the Claude models with the Claude Code harness; GPT-5.5 and GPT-5.6 Sol with the Codex harness.
- **SciCode.** Scores are cited from [Artificial Analysis](https://artificialanalysis.ai/) as of July 23, 2026.
- **Kimi Code Bench 2.0 (in-house).** Kimi K3 is evaluated with the Kimi Code harness (it attains 73.7 with the Claude Code harness); GLM-5.2, Claude Opus 4.8, and Claude Fable 5 with the Claude Code harness; GPT-5.5 and GPT-5.6 Sol with the Codex harness. All models are evaluated at maximum reasoning effort, except GPT-5.5, which uses the "xhigh" setting. As the benchmark includes cybersecurity and safety-related tasks, we also disclose the fraction of refused or fallback tasks: Claude Fable 5 hit 13 fallbacks and 1 refusal out of 80 tasks; 10 refusals out of 80 tasks entered GPT-5.6 Sol's cyber guard; GPT-5.5 had 3 refusals out of 80 tasks.
3. **Agentic benchmarks**
- **OfficeQA Pro.** Each test case provides the agent with the entire PDF corpus, with all PDFs rendered as images and no machine-readable text available.
- **OfficeQA Pro and SpreadsheetBench 2.** Kimi K3, GLM-5.2, Claude Opus 4.8, and Claude Fable 5 are evaluated with the Claude Code harness; GPT-5.5 and GPT-5.6 Sol are evaluated with the Codex harness.
- **MCP-Atlas.** All models are evaluated on the 500-task public subset with a 100-turn limit, using Gemini 3.1 Pro as the judge.
- **AutomationBench.** All models are evaluated on the 600-task public subset, following the official GitHub setup in all other respects.
- **BrowseComp.** We adopt a context-compaction strategy triggered at 300K tokens. When evaluated with the full 1M-token context window and no context management, Kimi K3 achieves a score of 90.4. The results of Claude Fable 5, Claude Opus 4.8, GPT-5.6 Sol, and GPT-5.5 are cited from [Anthropic](https://www.anthropic.com/news/claude-fable-5-mythos-5) and [OpenAI](https://openai.com/index/gpt-5-6/).
- **GDPval-AA v2, AA-Briefcase, ΟΒ³-Banking, Harvey Lab-AA, and APEX-Agents.** Scores are cited from [Artificial Analysis](https://artificialanalysis.ai/) and the [APEX-Agents leaderboard](https://www.mercor.com/apex/apex-agents-leaderboard/) as of July 23, 2026. For Harvey Lab-AA, we report the criterion pass rate.
- **CorpFin v2, Finance Agent v2, and Legal Research Bench.** Scores are cited from [Vals AI](https://www.vals.ai/).
- **Agents' Last Exam.** Scores are cited from the [official leaderboard](https://agents-last-exam.org/leaderboard) as of July 23, 2026; we report the leaderboard's primary pass-rate metric. On the leaderboard, each model is paired with a specific harness: Kimi K3 with Kimi Code; GPT-5.6 Sol and GPT-5.5 with Codex; Claude Fable 5, Claude Opus 4.8, and GLM-5.2 with Claude Code. <sup>β </sup> The Claude Fable 5 entry runs at xhigh effort with 40% of tasks annotated as downgraded.
4. **Multimodal benchmarks**
- Except for ZeroBench, which follows the official setting and is run five times, all multimodal scores are averaged over three runs. MMMU-Pro is evaluated following the official protocol, preserving the original input order and prepending images to the text input.
- **PerceptionBench** is an in-house benchmark that focuses on atomic visual perception capabilities.
</details>
## 4. Native MXFP4 Quantization
Kimi K3 applies quantization-aware training from the SFT stage onward, using MXFP4 weights with MXFP8 activations for broad hardware compatibility.
## 5. Deployment
> [!Note]
> You can access Kimi K3's API on https://platform.kimi.ai by selecting `kimi-k3`, and we provide OpenAI/Anthropic-compatible API for you. Currently, Kimi K3 is recommended to run on the following inference engines:
- [vLLM](https://github.com/vllm-project/vllm) β see [recipes](https://recipes.vllm.ai/moonshotai/Kimi-K3)
- [SGLang](https://github.com/sgl-project/sglang) β see [cookbook](https://docs.sglang.io/cookbook/autoregressive/Moonshotai/Kimi-K3)
- [TokenSpeed](https://github.com/lightseekorg/tokenspeed) β see [recipes](https://lightseek.org/tokenspeed/recipes/models#kimi-k3)
---
## 6. Model Usage
Kimi K3 always has thinking enabled, and will return `reasoning_content`. Thinking effort is configured with the top-level `reasoning_effort` request field, which supports `"low"`, `"high"`, and `"max"` (default `"max"`).
Kimi K3 was trained in the preserved thinking history mode. For multi-turn conversations and tool calls, Kimi K3 requires the complete assistant message returned by the API to be passed back to `messages` as-is β including `reasoning_content` and `tool_calls`, not just `content`:
```python
import openai
def chat_with_preserved_thinking(client: openai.OpenAI, model_name: str):
messages = [
{
"role": "user",
"content": "Tell me three random numbers."
},
{
"role": "assistant",
"reasoning_content": "I'll start by listing five numbers: 473, 921, 235, 215, 222, and I'll tell you the first three.",
"content": "473, 921, 235"
},
{
"role": "user",
"content": "What are the other two numbers you have in mind?"
}
]
response = client.chat.completions.create(
model=model_name,
messages=messages,
stream=False,
max_tokens=4096,
reasoning_effort="max",
)
# the assistant should mention 215 and 222 that appear in the prior reasoning content
print(f"response: {response.choices[0].message.reasoning}")
return response.choices[0].message.content
```
For full guides and examples (vision input, structured output, partial mode, tool choice, dynamic tool loading, context caching), see the [Kimi K3 Quickstart](https://platform.kimi.ai/docs/guide/kimi-k3-quickstart) and [Thinking Effort](https://platform.kimi.ai/docs/guide/use-thinking-effort).
### Coding Agent Framework
Kimi K3 works best with [Kimi Code CLI](https://www.kimi.com/code) as its agent framework. We warmly invite you to give it a try β run Kimi Code in your terminal and select Kimi K3 using the `/model` command. We hope you enjoy building with Kimi K3, and we would love to hear your feedback!
---
## 7. License
Both the code repository and the model weights are released under the [Kimi K3 License](https://huggingface.co/moonshotai/Kimi-K3/blob/main/LICENSE).
---
## 8. Contact Us
If you have any questions, please reach out at [support@moonshot.ai](mailto:support@moonshot.ai).
|