Upload folder using huggingface_hub

MiMo-V2.5-coder-Q2-00001-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:400d2928a0863f8e338717bdfc9d826cae7049ff968a79ac7b1220ebb93552f8
+oid sha256:2896b618eeb9d0e27bca57a3cf5ecd8520e19f4df296464c5db2fcf61b1a5adb
 size 8831584256

MiMo-V2.5-coder-Q2-00002-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c3820d01f8aac9c26b0430d4c3feef68fb5d02b8304418f5a994de6359b9ba94
+oid sha256:1d609955e1801579c62d6299cb49725696b8e6555f988aee10e9342c5d4c488f
 size 6996099872

MiMo-V2.5-coder-Q2-00003-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1043df380f9859dd1f42eea9862fbfd5987c737a679dc57941c8e6b52535a8bd
+oid sha256:a255ea8fce9fe2cde49203b938027927c96d3d0044ba9beb348f2792fc9c9a17
 size 6996099872

MiMo-V2.5-coder-Q2-00004-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6834f524a2f2de1153a11ba5f9509bd6268a240f891bf2dc4d311c52ee3f3e56
+oid sha256:404c6404f2b99d67a8b1a611003032e36e6047d25c1f9f93c73212afb92e638c
 size 6996099872

MiMo-V2.5-coder-Q2-00005-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8f50e2e0a565dc9dd67e45b6f1a6709bfe60e7f9918ef4e2e6f0ef300e43a98c
+oid sha256:b1623ed43c88610124cdb0e731356f0ff2f47d26e5d2a6d3793c7d93f0b2fcbb
 size 6996099872

MiMo-V2.5-coder-Q2-00006-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:88743a58cb0f6db6c96c26e8fdfc6973b2f73c4250d7b79b1cb11f4ff3f93999
+oid sha256:abc369764edb625ace3ad95ad522e8d2bfa0a5f3a4b6fc4759812be79e543011
 size 6996099872

MiMo-V2.5-coder-Q2-00007-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:472170a584ab7edbdf31bc8ccc8f998214de9aaf18355708c8df35b999c5f371
+oid sha256:7678e83adedf0903ae584173bb2b01e36a24e0faa6082246626a7bfdd315d8fb
 size 6996099872

MiMo-V2.5-coder-Q2-00008-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:22650e6f26ea15f6d76d467e910fe31ba46cc7423d83aaf73922d2599bac33a4
+oid sha256:5d85c683831f2b9942b817409b3967805a78e20d70598d43a642d6b12d0cccf2
 size 6996099872

MiMo-V2.5-coder-Q2-00009-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:df50089f1d5263d77e63ff6a6cc9856bf19b4d195cb6fc849eb109a3bbe867f5
+oid sha256:3d8493b982984c35a887f2a69839d5be3454a5574a96579b662f0cb1f94d5f58
 size 6996099872

MiMo-V2.5-coder-Q2-00010-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d6affc7d53669b35aca07ae7b057ddb3508120f5676115fca4022ead75899002
+oid sha256:89febfedaaaf5c5a2b576b9575594f6f732efb6e16895d025f16c3c3ac3bb877
 size 6996099872

MiMo-V2.5-coder-Q2-00011-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ef1be676dcfea5791849cf67086e14b2d3b756d4d40abcdac179ce7316f5297b
+oid sha256:59fbcc70d59edad135e92f310c70055bd09224b1369a119a084121c47f7426ee
 size 6996099872

MiMo-V2.5-coder-Q2-00012-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a14699f5394d85069574f49a93b6cc26643c3d3f6e7673e32bf756291ff02df2
+oid sha256:a8411d5242700a9acc72fb8eed3c1f7b9efe3d6cdf90091a7df7d6453f61786b
 size 6996099872

MiMo-V2.5-coder-Q2-00013-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b45d32e0266cb277a6fff36e19a8174d4118975a0ae7c5233fef9efdf0cbb97d
+oid sha256:d817b70895822965a6f929d9313376f5a46031d3165e88dccb637ec1c25faef8
 size 6996099872

MiMo-V2.5-coder-Q2-00014-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8b24c23aa55751b4a095ffa5e90165e6bd868a155885ed838878eea0963fa640
+oid sha256:a328343a4e68c901f2fe31d993df408680b38bcabb4d114a327b92c14027c074
 size 6996099840

MiMo-V2.5-coder-Q2-00015-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c22efbfcbd2c4287944b2cdc9dd3473430ce6ed247ad83118d8fb32c6ddf7ad3
+oid sha256:394281ac7b020ef0acd253a535b222aac02b4d488f979e864d63e12fdc2d7bee
 size 6996099872

MiMo-V2.5-coder-Q2-00016-of-00016.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:35051041037ab14d47c7e72502d8b19bf52832134c164acbb72b9e96bfba219e
+oid sha256:0de886b54240c049d747cad7264425f213b2e94e954cac2be18457808d7a98d1
 size 6996099872

README.md

@@ -9,6 +9,7 @@ tags:
 - llama.cpp
 - text-generation
 - code
+- coding
 - tool-calling
 - agent
 - mixture-of-experts
@@ -16,42 +17,123 @@ tags:
 pipeline_tag: text-generation
 ---
 
-# MiMo-V2.5 Coder Q2 GGUF
+# MiMo-V2.5 Coder Q2 v2 GGUF
 
-This is a local, self-quantized GGUF build of [XiaomiMiMo/MiMo-V2.5](https://huggingface.co/XiaomiMiMo/MiMo-V2.5), tuned for coding and tool-calling on a 128 GB Apple Silicon M5 machine.
+This is a text-only GGUF build of [XiaomiMiMo/MiMo-V2.5](https://huggingface.co/XiaomiMiMo/MiMo-V2.5), tuned for coding and OpenAI-compatible tool calling on high-memory local machines.
 
-This quant was optimized for systems with 128 GB of memory. The default serving profile targets a 128 GB Apple Silicon machine and tries to keep the model practical at a 100,000-token context. Smaller-memory systems will likely need more aggressive CPU offload, a smaller context, or a different quant.
+The target system for this build is a 128 GB Apple Silicon machine. The default serving profile uses a 100,000-token context and asks llama.cpp to fit as much of the model as possible onto Metal while leaving enough headroom for KV cache and runtime buffers. Smaller-memory machines will likely need a smaller context, more CPU offload, or a smaller quant.
 
-It is a text-only quantization. The original MiMo-V2.5 checkpoint is omnimodal, but this GGUF does not include the vision or audio encoders. The MiMo multi-token prediction blocks were also omitted during conversion because normal llama.cpp generation does not currently execute them for this model.
+This is not a multimodal build. MiMo-V2.5 is an omnimodal checkpoint, but this GGUF contains the text model only. The vision and audio encoders are not included. MiMo's multi-token prediction blocks were also omitted because the current llama.cpp MiMo2 generation path does not use those blocks for normal inference.
 
-## Quantization
+## Why this build exists
 
-High-level summary:
+Public low-bit quants of very large MoE models can be surprisingly fragile: tool calls may become malformed, code may fail on small API details, and long answers can drift into repeated reasoning loops. This build was made to spend the limited Q2-class quality budget on the workloads where MiMo-V2.5 is most useful locally:
+
+- coding in common systems and scripting languages
+- web UI/component generation
+- OpenAI-compatible tool calling
+- Swival-style agent loops over real files and commands
+- long English technical prompts
+
+Chinese-language quality and multimodal behavior were not optimization targets.
+
+## How it was built
+
+The source was the original `XiaomiMiMo/MiMo-V2.5` checkpoint, converted to GGUF with llama.cpp's native MiMo2 support. Conversion was text-only and omitted runtime-inactive MTP/NextN blocks, so memory is not spent on tensors that current llama.cpp MiMo2 inference does not execute.
+
+The final artifact is a split `Q2_K_S` GGUF with an importance matrix built from English coding, debugging, tool-calling, shell, and agent prompts. The calibration mix was designed to make the quantizer preserve behavior that matters for developer workflows rather than generic chat breadth.
+
+The build was iterative:
+
+1. Convert the original checkpoint to split BF16 GGUF.
+2. Produce a first low-bit coding/tool-use candidate.
+3. Test that candidate on executable coding tasks and Swival-style tool calls.
+4. Add calibration coverage for the failures that showed up in real tests.
+5. Rebuild the importance matrix from the expanded coding/tool-use prompt mix.
+6. Re-quantize with the final `Q2_K_S` recipe.
+
+The calibration text is not required to use the model. It was a build-time tool for telling the quantizer which activations mattered most: code generation, code repair, shell-style work, JSON/tool-call formatting, and agent workflows over real files.
+
+Quantization details:
 
 - Quant type: `Q2_K_S`
 - Importance matrix: coding and tool-calling focused
-- Preserved higher precision for embeddings, output, attention, and the dense first FFN
-- MoE down-expert tensors: `Q3_K`
-- Reported quantized size: about 108,496.76 MiB at 2.95 BPW
+- Embeddings and output tensors kept at higher precision
+- Attention and dense first-FFN tensors protected at higher precision
+- MoE down-expert tensors kept at `Q3_K`
+- Reported size: about 108,496.76 MiB, 2.95 BPW
+- Split files: 16 GGUF shards
+
+One tokenizer metadata fix is included: the base-vocabulary `</s>` token is marked as a control-looking token so llama.cpp does not warn at load time. MiMo's real EOS token remains `<|im_end|>`.
+
+## Why this recipe was chosen
+
+The recipe is a compromise between quality and a hard practical limit: this model has to run locally on a 128 GB unified-memory machine. Higher-bit GGUFs of a model this large can exceed the useful memory envelope once KV cache, batching, Metal buffers, and the operating system are included.
+
+The first plain `Q2_K` family candidate was small enough, but it was not reliable enough for tool calling. It malformed some Swival-style arguments and missed several conditional tools. The v2 recipe is larger, but it spends the extra space where it helped most:
+
+- embeddings and output tensors stay higher precision because they are important for token identity and exact syntax
+- attention tensors are protected because tool-call and code prompts are structure-heavy
+- the dense first FFN is protected because early-layer representation quality matters disproportionately after heavy quantization
+- MoE down-expert tensors use `Q3_K`, which was a better quality/memory tradeoff than pushing all expert down-projections lower
+
+That is why this is still a Q2-class build, but not the smallest possible Q2 build.
+
+## Why it is good at coding
+
+This quant was not chosen just because it fits in memory. It was iterated against executable tasks and then rebuilt with a stronger coding/tool-use importance matrix after early failures were identified.
+
+The first low-bit pass exposed the kinds of issues that matter in practice: malformed tool-call arguments, brittle JavaScript Markdown parsing, incorrect Zig checked-addition APIs, and small C/C++/Go harness problems. Those failures were used to improve the calibration distribution and to validate that the final model can solve the tasks when the problem statement contains the same constraints a developer would normally give.
+
+The final v2 artifact passed the local coding and web-design harness across:
 
-One tokenizer metadata fix is included so llama.cpp does not warn about the base-vocab `</s>` token at load time. MiMo's actual EOS token remains `<|im_end|>`.
+- Swift
+- JavaScript
+- TypeScript through Deno
+- Rust
+- C
+- C++
+- Zig
+- Python
+- Perl
+- Go
+- static HTML/CSS
 
-## Imatrix
+That harness writes complete model-generated files into isolated directories and validates them with local compilers, runtimes, or test runners. The current v2 run passed 11/11. The checks are intentionally practical rather than benchmark-like: they catch whether the generated code compiles, runs, and handles edge cases from the prompt.
 
-This build deliberately prioritizes:
+It was also tested on framework-style frontend tasks. React, Vue, and Solid components were rendered server-side with Deno/npm tooling, including props, filtering behavior, accessible form markup, and summary text checks. The current v2 run passed 3/3.
 
-- reliable OpenAI-compatible tool calls
-- coding and shell-oriented agent use
-- English prompts and codebase work
-- practical inference on a 128 GB Apple Silicon system
+The important point is not that these small harnesses prove universal coding ability. They prove that the quantization process did not destroy the details that low-bit models often lose first: exact exported names, balanced parsing logic, checked arithmetic APIs, command/tool argument shapes, and framework-specific rendering conventions.
 
-Chinese-language quality and multimodal use were not optimization targets.
+## Tool-calling validation
 
-## Serving
+Tool calling was tested with Swival because Swival exercises OpenAI-style tools in realistic agent loops instead of only checking toy single-call examples.
+
+Validation included:
+
+- a broad synthetic selector suite covering the current Swival tool surface
+- real one-shot Swival tasks over files, grep, command execution, fetches, image input, skills, metaskills, snapshots, todos, and subagents
+- a real `/goal` run that required the model to complete work and call `complete_goal`
+
+The current v2 results were:
+
+- Swival all-tools selector: 22/22
+- real Swival one-shot suite: 10/10 with zero failed tool calls
+- real Swival goal-mode `complete_goal`: passed with exactly one successful `complete_goal` call
+
+A separate repetition-loop guard was also run on long coding and web prompts. The current v2 artifact passed 4/4, with no repeated-tail failures.
+
+These are local validation results, not public benchmark scores. They are included so users know what this quant was optimized for and what kinds of regressions were actively checked.
+
+Compared with the earlier local candidate, the v2 build fixed the key practical failures: the broad Swival selector went from 18/22 to 22/22, the coding/web suite reached 11/11 after task prompts were aligned with the validators, and the real Swival task suite completed with zero failed tool calls. This is why the package is labeled `v2`.
+
+## Serving with llama.cpp
+
+Recent llama.cpp builds should be able to load the repo directly:
 
 ```sh
 llama-server \
-  -hf jedisct1/MiMo-V2.5-coder-Q2 \
+  -hf jedisct1/MiMo-V2.5-coder-Q2-v2 \
   --host 127.0.0.1 \
   --port 8080 \
   --ctx-size 100000 \
@@ -73,17 +155,27 @@ llama-server \
   --reasoning off
 ```
 
-This starts an OpenAI-compatible server on `127.0.0.1:8080`. The repository contains one GGUF split set, so recent llama.cpp builds should select the first shard automatically.
+If your llama.cpp build does not auto-select the split GGUF set, pass the first shard explicitly:
 
-If you cloned or downloaded the repository locally, you can use the included helper script instead:
+```sh
+llama-server \
+  -hf jedisct1/MiMo-V2.5-coder-Q2-v2 \
+  --hf-file MiMo-V2.5-coder-Q2-00001-of-00016.gguf \
+  --ctx-size 100000 \
+  --flash-attn on \
+  --jinja \
+  --reasoning off
+```
+
+If you cloned or downloaded the repository locally, you can use the helper script:
 
 ```sh
 ./run-server.sh
 ```
 
-The script uses the same defaults and loads the first GGUF shard from the repository directory.
+The helper script loads the first GGUF shard next to it and uses the same default serving profile.
 
-Default serving settings:
+Default settings:
 
 ```sh
 MIMO_CTX=100000
@@ -95,77 +187,28 @@ MIMO_REASONING=off
 MIMO_CPU_MOE=0
 ```
 
-These defaults are tuned for an Apple M5 Max with 128 GB unified memory. They keep reasoning output disabled, use the model's Jinja chat template, use Flash Attention, and ask llama.cpp to fit as much of the model as possible onto Metal.
-
-If you hit memory pressure, use the safer CPU-MoE mode:
+For more memory headroom, use CPU-MoE mode:
 
 ```sh
 MIMO_CPU_MOE=1 MIMO_FIT_TARGET=32768 MIMO_BATCH=128 MIMO_UBATCH=64 ./run-server.sh
 ```
 
-That mode is slower, especially on long prompt prefill, but it leaves much more Metal memory headroom.
+That mode is slower, especially during long prompt prefill, but it leaves more Metal memory available.
 
-You can point directly at a different `llama-server` binary with:
+You can point the script at a specific server binary:
 
 ```sh
 LLAMA_SERVER=/path/to/llama-server ./run-server.sh
 ```
 
-You can also run `llama-server` directly against local files without the helper script:
-
-```sh
-llama-server \
-  --model MiMo-V2.5-coder-Q2-00001-of-00016.gguf \
-  --host 127.0.0.1 \
-  --port 8080 \
-  --ctx-size 100000 \
-  --parallel 1 \
-  --batch-size 512 \
-  --ubatch-size 128 \
-  --threads 12 \
-  --threads-batch 18 \
-  --prio 0 \
-  --poll 80 \
-  --flash-attn on \
-  --jinja \
-  --fit on \
-  --fit-target 4096 \
-  --fit-ctx 100000 \
-  --gpu-layers auto \
-  --cache-type-k f16 \
-  --cache-type-v f16 \
-  --reasoning off
-```
-
-For the safer CPU-MoE fallback, add `--cpu-moe` and use a larger fit margin:
-
-```sh
-llama-server \
-  --model MiMo-V2.5-coder-Q2-00001-of-00016.gguf \
-  --ctx-size 100000 \
-  --fit on \
-  --fit-target 32768 \
-  --fit-ctx 100000 \
-  --batch-size 128 \
-  --ubatch-size 64 \
-  --flash-attn on \
-  --jinja \
-  --gpu-layers auto \
-  --cache-type-k f16 \
-  --cache-type-v f16 \
-  --reasoning off \
-  --cpu-moe
-```
-
-## Tool-Calling Notes
-
-For best tool-calling results:
+## Tool-calling tips
 
-- Use the [Swival](https://swival.dev) harness - it should work with anything using OpenAI-like tool calling convention, but it is tested with Swival.
-- Disable model reasoning output with `--reasoning off` or `MIMO_REASONING=off`.
+- Disable reasoning output with `--reasoning off` or `MIMO_REASONING=off`.
+- Send tool schemas from the client rather than enabling llama.cpp built-in tools.
 - Set `parallel_tool_calls` to `false` if your client supports it.
-- Avoid forcing `tool_choice: required`; in testing it made malformed calls more likely.
+- Avoid forcing `tool_choice: required`; in testing, that made malformed calls more likely.
+- Use a client that supports OpenAI-compatible tool calls cleanly. Swival was the main validation client for this build.
 
 ## License
 
-The upstream model card for `XiaomiMiMo/MiMo-V2.5` declares the MIT license. This derived GGUF is provided under the same license metadata.
+The upstream `XiaomiMiMo/MiMo-V2.5` model card declares the MIT license. This derived GGUF is provided with the same license metadata.