Xylolabs

TML-d24-varloop-research GGUF

Base model License Code

These llama.cpp GGUFs of TML-d24-varloop-research include a separate vision projector and omit the MTP drafter.

Files

File Size
TML-d24-varloop-research-NVFP4.gguf 22.92 GB
TML-d24-varloop-research-NVFP4-Q8_0.gguf 17.63 GB
mmproj-TML-d24-varloop-research-BF16.gguf 641.81 MB
mmproj-TML-d24-varloop-research-Q8_0.gguf 345.91 MB

How to run

Use llama.cpp release b12611 or later, with the Hugging Face CLI to download the files. Earlier builds skip the per-expert NVFP4 scales on the second and third loop passes and give wrong output with these files.

hf download Xylolabs/TML-d24-varloop-research-GGUF TML-d24-varloop-research-NVFP4-Q8_0.gguf mmproj-TML-d24-varloop-research-BF16.gguf --local-dir .
llama-server -m TML-d24-varloop-research-NVFP4-Q8_0.gguf --mmproj mmproj-TML-d24-varloop-research-BF16.gguf --jinja

For text-only use, omit --mmproj and the projector filename.

The default loop count is three. To change it at load time, pass --override-kv inkling2_loop.loop.count=int:N with the desired count in place of N.

Accuracy

Mean KL divergence from the BF16 reference (lower is better) and top-1 agreement with it, measured with llama.cpp on 512-token chunks.

Quant Size WikiText-2 KL top-1 Chat KL top-1 Held-out KL top-1
NVFP4 22.92 GB 0.0026 97.8% 0.0142 96.9% 0.0025 97.8%
NVFP4-Q8_0 17.63 GB 0.0034 97.6% 0.0194 96.4% 0.0030 97.4%
ggml-org IQ4_NL 17.63 GB 0.0098 96.0% 0.0307 95.0% 0.0069 96.4%
ggml-org Q8_0 27.82 GB 0.0033 97.5% 0.0169 96.5% 0.0029 97.7%
llama.cpp baseline Q4_K_M 17.48 GB 0.0188 94.2% 0.0449 93.7% 0.0137 94.9%
WikiText-2 mean KL divergence by text GGUF size WikiText-2 top-1 agreement by text GGUF size

Which file for your hardware

File size, plus the KV cache at the stated context (f16), plus about 2 GB of runtime overhead, must fit in 90% of the memory. Its layers 6-17 run three times, each pass with its own cache: 8 caches use full attention (8 KV heads, head size 128) and 40 use a 1,024-token sliding window, a fixed cache of about 0.25 GB.

Hardware Memory Recommended file Context that fits Notes
RTX 4090 / RTX 3090 24 GB TML-d24-varloop-research-NVFP4-Q8_0.gguf (17.6 GB) 32K
RTX 5090 32 GB TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 64K
RTX 6000 Ada, or 2 x 24 GB 48 GB TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 512K Two 24 GB cards need the layers split across them.
One H100 80 GB TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 1M
RTX PRO 6000 Blackwell (GDDR7) 96 GB TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 1M
DGX Spark 128 GB unified TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 1M
Mac Studio M5 Max 128 GB unified TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 1M The GPU may use about 75% of unified memory by default; sudo sysctl iogpu.wired_limit_mb=N raises it.
One H200 (HBM3e) 141 GB TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 1M
One B200 (about 180 GB usable in a DGX B200) 192 GB TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 1M
Mac Studio M5 Ultra 256 GB unified TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 1M The GPU may use about 75% of unified memory by default; sudo sysctl iogpu.wired_limit_mb=N raises it.
Mac Studio M5 Ultra 512 GB unified TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 1M The GPU may use about 75% of unified memory by default; sudo sysctl iogpu.wired_limit_mb=N raises it.
One B300 (HBM3e; usable a little less) 288 GB TML-d24-varloop-research-NVFP4.gguf (22.9 GB) 1M

License and notices

These GGUF editions are licensed under Apache 2.0 and subject to Thinking Machinesโ€™ Model Acceptable Use Policy. The NOTICE identifies the copyright holder and Xylolabsโ€™ changes. Xylolabs is independent of Thinking Machines Lab.

Updates

  • 2026-11-22: Added a hardware recommendation table.
  • 2026-11-22: llama.cpp release b12611 or newer runs these files; earlier builds give wrong output.
  • 2026-11-21: First release.

Thinking Machines' original model card

Thinking Machines Labโ€™s original model card follows.

TML-d24-varloop-research

1. General Information

TML-d24-varloop-research is a multimodal research model that accepts text and image inputs and generates text outputs. It is a variable-looped counterpart to Inkling 2 Mini: whereas Inkling 2 Mini always runs its looped layers a fixed number of times, the number of loops in this model can be chosen at inference time. We are releasing it with open weights to support research on looped models. Because the number of loops can be varied on a single set of weights, the model allows cleaner comparisons of how additional looped computation affects a model's internal computation and overall capabilities.

Languages: English, general multilingual support

Model provider: Thinking Machines Lab, Inc.

Date of release: Nov 20, 2026

License: Apache 2.0

Model Acceptable Use Policy

2. Getting Started

The weights are available for download through Hugging Face here.

The model is not available through an API, including Tinker. It is distributed as open weights only.

Required hardware

The model is distributed as a single checkpoint of about 20GB. Its expert layers are NVFP4, the last four use MXFP8, and the remaining weights are BF16.

The checkpoint requires a single machine with at least 32GB of GPU or unified memory. This can be met with any of the following example configurations:

  • 1x NVIDIA DGX Spark
  • 1x NVIDIA GPU with 32GB or more of VRAM (for example, GeForce RTX 5090 or RTX PRO 6000 Blackwell)
  • Apple silicon Mac with 32GB or more of unified memory (for example, Mac mini)

The model uses the same mixed NVFP4 and MXFP8 precision it was trained in, so this checkpoint is its native precision and needs no further quantization.

Required software

TML-d24-varloop-research supports local inference using the following frameworks, along with their dependency libraries:

TML-d24-varloop-research can also be post-trained with:

3. Model Properties

Model type

Multimodal autoregressive looped transformer

Architecture type

A 24-layer decoder-only transformer with sparse Mixture-of-Experts (MoE) feed-forward layers. The middle 12 layers run a variable number of times with shared weights, chosen at inference time; the model was trained with 1 to 5 passes. In each layer, each token is routed to 8 of 128 experts. Attention is a hybrid of local and global layers. The model is natively multimodal โ€“ images are encoded via a vision transformer โ€“ with image and text inputs projected into a shared hidden space and processed jointly by the decoder.

Parameters

26.4B, 3.4B active

Note: each extra pass through the looped layers adds about 1.0B effective active parameters per token. This corresponds to 3.4B active parameters at 1 pass, 5.4B at 3 passes, and 7.4B at 5 passes.

Numerics support

NVFP4 and MXFP8

Input modalities

  • Text: UTF-8 encoded text
  • Image: Any pixel-based image input. For optimal performance, each image dimension should be between 40px to 4096px.

TML-d24-varloop-research supports a context window of up to 1M tokens.

Output modalities

Text

4. Training

Training data includes a broad variety of content types, including text, images, video. Training data for the model was drawn from publicly available sources, acquired from third-parties, generated internally, or synthetically generated or augmented. Publicly available data includes content from the public internet and publicly accessible repositories.

Data curation includes cleaning, processing, and modifying datasets. These processing steps, which vary by data type, may include deduplication and filtering to remove junk or other low-quality data, or to advance safety or other objectives.

Training with variable looping

TML-d24-varloop-research shares its pretraining and first stage of midtraining with Inkling 2 Mini. For the rest of midtraining, it was trained with 1 to 5 passes through the looped layers, sampled in a 1:2:6:2:1 ratio, and it was then post-trained with 3 passes. Because 3 passes dominated training once looping was introduced, performance at other pass counts may understate what models trained specifically at those counts could achieve. The two models' later training stages differ, so we caution against comparing TML-d24-varloop-research directly to Inkling 2 Mini.

5. Evaluations

Each row shows one benchmark score for TML-d24-varloop-research run at 1 to 5 passes through its looped layers, beside Inkling 2 Mini, which always runs 3 passes. Scores are in percent; โ€“ marks a pass count without a completed result yet.

Inkling 2 MiniTML-d24-varloop-research
(3 passes)1 pass2 passes3 passes4 passes5 passes
Agentic Coding
DeepSWE v1.1
mini-swe-agent v2.4.6
48.2%0.7%28.3%47.4%46.9%48.2%
SWEBench Pro v2 (Hard)
mini-swe-agent v2.4.6
77.9%39.7%69.1%78.9%76.0%78.4%
SWEBench Multilingual
bash only
73.7%40.8%68.5%71.8%72.8%72.0%
SWEBench Verified
bash only
79.2%56.3%76.5%78.7%79.3%79.6%
Agentic Search, Tool Use, and Knowledge Work
BrowseComp
original, no context management
76.0%29.4%53.8%65.7%69.4%71.0%
MCP Atlas
Public
82.7%70.0%77.6%79.4%79.7%80.0%
JobBench49.6%22.8%55.1%63.8%61.8%63.3%
EnterpriseOps-Gym
LangChain ReAct loop
36.7%25.7%34.4%36.6%36.5%37.6%
Vision Interpretation and Creation
GDP.pdf16.8%4.6%11.2%19.6%18.0%20.0%
PerceptionBench57.5%45.5%51.3%53.9%53.7%52.7%
MMMU Pro
Standard 10
74.9%63.6%69.8%71.4%72.1%73.2%
OmniDocBench v1.692.8%86.6%90.5%92.1%91.8%92.5%
BabyVision
original
53.4%32.7%44.9%50.3%50.3%49.2%
BabyVision
with Python
88.8%67.2%84.1%88.9%88.8%89.5%
ScreenSpot Pro
original
76.5%54.4%66.0%72.0%73.4%73.1%
ScreenSpot Pro
with Python
87.7%80.7%86.5%87.7%87.2%88.1%
Charxiv RQ
original
80.6%75.0%78.3%79.9%79.2%80.4%
Charxiv RQ
with Python
88.3%84.7%87.7%87.6%89.3%87.9%
Chartography
original
22.1%14.5%17.8%21.6%21.4%22.6%
Chartography
with Python
72.5%49.1%69.1%72.3%73.7%72.3%
ZeroBench (main)
original
17.0%9.0%13.0%19.0%18.0%17.0%
ZeroBench (main)
with Python
34.0%16.0%31.0%40.0%35.0%37.0%
Humanityโ€™s Sixth Sense23.8%22.4%24.8%25.5%26.5%27.0%
MathVision89.0%74.8%84.9%86.6%87.8%88.1%
MuirBench77.5%73.6%76.1%77.1%76.4%77.8%
BVB68.4%53.9%64.4%68.5%69.2%69.5%
Reasoning, Knowledge, and Epistemics
HLE-Diamond15.2%6.3%8.5%9.5%9.7%11.1%
HLE (text only)30.7%10.5%20.0%24.9%25.4%25.9%
GPQA Diamond86.6%69.6%79.2%84.2%81.9%83.1%
Apex 202534.4%3.1%10.4%19.8%15.6%18.8%
ComplexConstraints v1.1
openhands v1.45.0
33.4%4.5%26.0%37.9%39.6%38.9%
SimpleQA Verified8.4%6.3%6.5%7.3%6.8%6.7%
AA-Omniscience v1.1 (Public)
accuracy
21.3%15.5%16.3%18.4%18.0%17.4%
AA-Omniscience v1.1 (Public)
non-hallucination rate
64.6%79.3%85.0%84.0%85.7%85.9%
SimpleVQA45.4%39.0%41.8%42.4%42.9%41.3%
WorldVQA15.2%11.3%13.0%14.2%14.3%13.5%
GraphWalks BFS
256K
69.7%23.5%56.0%64.3%64.3%63.8%
MRCR v2
8-needle 256Kโ€“512K
46.4%27.2%65.2%89.2%93.4%92.1%
Safety and Alignment
FORTRESS (adversarial)
Public
89.5%89.0%90.4%90.0%89.9%89.4%
FORTRESS (benign)
Public
96.4%95.0%95.9%95.3%95.7%95.0%
StrongREJECT (none)97.8%98.8%98.4%98.2%99.4%98.4%

6. Safety

We evaluated this model with the same pre-release process as Inkling 2 Mini. It is less capable than models whose weights we have already released on CBRN and cyber evaluations, and it refuses harmful requests about as reliably as Inkling 2 Small.

For production use, we recommend Inkling 2 Mini as it generally performs better on capabilities and alignment evaluations. We do not intend TML-d24-varloop-research for production use.

Across all areas, we concluded that TML-d24-varloop-research did not present risk of material uplift beyond whatโ€™s already available in the open-weight ecosystem.

7. Bias, risks and limitations

TML-d24-varloop-research is a research model, not a production model. It may exhibit general limitations common to foundation models, including hallucination (generating plausible but factually incorrect or unsupported content), occasional failures to follow instructions precisely, and degraded performance in long multi-turn conversations. As with other large-scale models trained on web-derived and synthetic data, TML-d24-varloop-research may reflect biases present in its training data, including demographic, cultural, or linguistic biases, and may perform unevenly across languages, dialects, or subject domains that were less represented during training.

TML-d24-varloop-researchโ€™s knowledge is limited to information available as of its training cutoff, and it may not reflect events, developments, or changes that occurred afterward.

TML-d24-varloop-research is intended to be a research artifact. It has not been carefully tuned for deployment, and may behave unpredictably at pass counts other than 3. We do not recommend deploying this model in production or user-facing applications.

8. Legal

Training Data Documentation

Downloads last month
439
GGUF
Model size
26B params
Architecture
inkling2_loop
Hardware compatibility
Log In to add your hardware

4-bit

8-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for Xylolabs/TML-d24-varloop-research-GGUF

Quantized
(4)
this model

Collection including Xylolabs/TML-d24-varloop-research-GGUF