whatmodel.sh / Mellum2.1 12B-A2.5B Thinking
Mellum2.1 12B-A2.5B Thinking
↗ share
+ contribute
Mellum2.1 12B-A2.5B Thinking is an open-weights language model from JetBrains, released September 2026. Its published size is 12.15B parameters (2.5B active per token). Licence: apache-2.0.
Catalogue identity notes
JetBrains/Mellum2.1-12B-A2.5B-Thinking @92ddae9 (created 2026-09-20, licence apache-2.0). JetBrains Mellum2.1 thinking model (Sept 2026), successor of Mellum2 Thinking. Rows from its own card table (JetBrains' runs of Mellum2 Thinking, Gemma 4 E4B and Qwen3.5 9B in thinking mode).
Model information sources
Release date ↗ · Weights licence ↗
HF safetensors metadata 12,149,923,072 parameters; 2.5B active per the maker card.
$ how it compares
The LLMs ranked either side of Mellum2.1 12B-A2.5B Thinking, out of 828 with a score, then more from JetBrains.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 226 |  Qwen3.5 4B | 61±1 | local | ||
| 227 |  Nemotron-Cascade-14B-Thinking | 61±1 | local | ||
| 228 | Mellum2.1 12B-A2.5B Thinking | 60±2 | local | ||
| 229 |  GPT-4.1 mini | 60±1 | api $0.70/M | ||
| 230 |  DeepSeek V3 0324 | 60±1 | local | ||
| more from JetBrains | |||||
| 340 | Mellum2 12B-A2.5B Thinking | 54±1 | local | ||
| 410 | Mellum2 12B-A2.5B Instruct | 50±2 | local | ||
$ what it's good at
agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #144 of 300 · 35% of its score · 4 tests
Beats Qwen3 235B-A22B Thinking 2507, o4-mini, o1 (2024-12-17) here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, Jamba 1.5 Large here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Terminal-Bench 2.1 lab | 17.4% ✎ | ▁·▁·▁▃▄▃▄█▃▇ beats 3/56 | 12% |
| SWE-bench Verified lab | 47.0% ✎ | ▁▁▁▁▁▁▂▄█▆▁▁ beats 36/131 | 8.7% |
| SWE-bench Pro (split unspecified) lab | 28.0% ✎ | ▃·▁▁▁▂▃▃█▇▃▁ beats 13/55 | 8.3% |
| BFCL V4 (December 2025 archive) lab | 62.3% ✎ | ▁▁▁▂▇▆▆▆█▃▄▂ beats 95/111 | 6.1% |
coding▁▁▁▃▆██▇▆▄▃▁ #166 of 458 · 40% of its score · 5 tests
Beats Qwen3-Coder-Next (80B-A3B), Qwen3-VL 235B-A22B Instruct, Qwen3-Next 80B-A3B Instruct here, though they rank higher overall. Trails QwQ-32B-Preview, Ministral 3 8B Reasoning 2512, Phi-4-reasoning 14B here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LiveCodeBench v6 (window unspecified) lab | 82.0% ✎ | ▂▃▄▄▅██▅▄▇▃▆ beats 47/65 | 11% |
| SWE-bench Verified lab | 47.0% ✎ | ▁▁▁▁▁▁▂▄█▆▁▁ beats 36/131 | 8.7% |
| SWE-bench Pro (split unspecified) lab | 28.0% ✎ | ▃·▁▁▁▂▃▃█▇▃▁ beats 13/55 | 8.3% |
| HumanEval+ lab | 91.5% ✎ | ▁·▁▁▂▃▄▄█▅▄▃ beats 136/145 | 7.9% |
| MBPP+ lab | 79.4% ✎ | ▁··▁▁▃▄█▆▄▁▁ beats 108/120 | 4.3% |
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #165 of 393 * · 3.4% of its score · 1 test
Beats MiniMax-M2, MiniCPM5 2B, Granite 4.2 30B here, though they rank higher overall. Trails LFM2.5 1.2B Instruct, LFM2.5 1.2B Thinking, LFM2.5 8B-A1B here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| IFEval (aggregation unspecified) lab | 90.6% ✎ | ▁▁▃▁▂▄▄▆█▃▂▁ beats 118/141 | 3.4% |
other tests3 not in a task · 38% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond lab | 64.6% ✎ | ▁▁▁▁▂▄▅▅███▄ beats 128/310 | 26% |
| MMLU-Redux lab | 87.8% ✎ | ▁▁▃▁▂▃▄▅▇▅█▄ beats 27/63 | 8.8% |
| GSM-Plus lab | 88.3% ✎ | ▁··▁▁▁·▃▁▄█▁ beats 14/16 | 2.8% |
$ where to run it
llama.cpp · recommended
version
- Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.macOS / Linux
curl -LsSf https://llama.app/install.sh | shWindowsirm https://llama.app/install.ps1 | iex - Run Mellum2.1 12B-A2.5B Thinking · needs ~8.1 GB · from huggingface.co/JetBrains (official)chat in the terminal
llama cli -hf JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF:Q4_K_M - Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF:Q4_K_M
LM Studio
version
- Install LM Studio · Open the app once before using the lms command.
- Run Mellum2.1 12B-A2.5B Thinking · needs ~8.1 GB · from huggingface.co/JetBrains (official)
- Or download it in the terminal
lms get https://huggingface.co/JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF@Q4_K_M -y - Then run lms ls to see its name, and lms chat followed by that name
Ollama
version
- Install Ollama
download it (macOS / Windows) ↗
Linuxcurl -fsSL https://ollama.com/install.sh | sh - Run Mellum2.1 12B-A2.5B Thinking · needs ~8.1 GB · from huggingface.co/JetBrains (official)chat in the terminal
ollama run hf.co/JetBrains/Mellum2.1-12B-A2.5B-Thinking-GGUF:Q4_K_M
MLX (Mac)
- Install MLX (Mac) · Apple silicon Macs only, text only.with uv
uv tool install mlx-lmorpip install mlx-lm - Run Mellum2.1 12B-A2.5B Thinking · needs ~6.5 GB, Mac only · from huggingface.co/randmaruchat in the terminal
mlx_lm.chat --model randmaru/Mellum2.1-12B-A2.5B-Thinking-mlx-mxfp4 - Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model randmaru/Mellum2.1-12B-A2.5B-Thinking-mlx-mxfp4
Other ways to run it
- Run it as published · needs ~24.3 GB, NVIDIA only · from huggingface.co/JetBrains (official)
vllm serve JetBrains/Mellum2.1-12B-A2.5B-Thinking
Official weights and documentation: huggingface.co/JetBrains ↗
$ how the score was worked out · JetBrains/Mellum2.1-12B-A2.5B… 99%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 JetBrains/Mellum2.1-12B-A2.5B-Thinking model card · lab-reported | 11 | 99% |
| prior · pull toward the middle | 1% | |
| weighted average | = 60.4 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json