Mellum2.1 12B-A2.5B Thinking
12B-A2.5BJetBrains Mellum
JetBrains' open coding model for agents and fast sub-agents, released in October under Apache 2.0. It is a 12B mixture-of-experts model with about 2.5B parameters active per token, so it generates at roughly the speed of a 3B model. At Q4_K_M it needs about 8.6 GB at 4K context and 10.5 GB even at its full 128K window, because only 7 of its 28 layers keep a growing cache. That is comfortable on a 10 or 12 GB card; an 8 GB card cannot hold it. Q8_0 fits a 16 GB card. llama.cpp supports it as `mellum`. Sizes come from JetBrains' own GGUF files; JetBrains publishes KL divergence rather than perplexity, so no quality-loss figure is shown.
131K
Max Context
2
Quant Variants
GGUF Q8_0
Best Quality
—
Accuracy Retained
Quantization Variants
Per-quant VRAM, quality loss, and inference speed on RTX 4090
Measured = run on this site · Estimated = not run here: calculated, or a figure this site has not verified · Community = public reports
Similar models
Compare with Qwen2.5 14BQwen2.5 14B Instruct
Alibaba Qwen2.5
The sweet spot between performance and resource usage. 16GB VRAM with Q4.
DeepSeek-Coder-V2-Lite Instruct
DeepSeek
MoE architecture coding model. Active params ~2.4B, total ~16B. Exceptional code quality. Its attention is MLA: a GGUF converted by current llama.cpp (tensors attn_k_b / attn_v_b) caches about 1 GB at 32K, which is what the estimates here assume. An older conversion with a single attn_kv_b tensor caches full keys and values instead — about 8.4 GB at 32K.
Phi-3 Medium 14B Instruct
Microsoft Phi
Microsoft's mid-size Phi-3. Excellent quality-per-GB on 16GB cards.
DeepSeek-V2-Lite Chat
DeepSeek
MoE general model (~2.4B active). Long context and strong multilingual chat. Its attention is MLA: a GGUF converted by current llama.cpp (tensors attn_k_b / attn_v_b) caches about 1 GB at 32K, which is what the estimates here assume. An older conversion with a single attn_kv_b tensor caches full keys and values instead — about 8.4 GB at 32K.
Running Mellum2.1 12B-A2.5B Thinking locally
At Q4_K_M and 4K of context, Mellum2.1 12B-A2.5B Thinking needs about 8.6 GB — 7.7 GB of weights, 0.12 GB of KV cache and a 0.8 GB activation buffer. The smallest cards in this index that clear that comfortably are 10 GB ones — 2 of them, from the Arc B570 10G up — and 67 of the 78 cards here do. These are calculated figures, not measurements: the estimate stops counting a card as comfortable at 88% of its VRAM, which is roughly the room a desktop session needs.
What longer context costs
Going from 4K to 32K adds about 0.38 GB, taking the total to 9.0 GB. That is gentler than the parameter count suggests: only 7 of 28 layers keep a KV cache that grows with context, and 21 more cap theirs at a sliding window, so the cache scales at a fraction of the usual rate. The model's native window is 128K; holding all of it at Q4_K_M would need about 10 GB.
Which build to download
This index only tracks Mellum2.1 12B-A2.5B Thinking in GGUF, across 2 levels (Q4_K_M, Q8_0). No per-level perplexity sweep has been published for this model, so the quality column is empty rather than estimated — within one model, more bits per weight is the only ordering the data supports. No throughput has been measured for this model on this site's hardware. GGUF runs on llama.cpp and Ollama across NVIDIA, AMD and Apple silicon; AWQ and GPTQ target vLLM on CUDA and ROCm; EXL2 is ExLlamaV2 — CUDA only, and now archived, so local use rather than serving.
Common questions
- How much VRAM does Mellum2.1 12B-A2.5B Thinking need?
- About 8.6 GB at Q4_K_M with 4K of context and batch 1, rising to roughly 9.0 GB at 32K. That figure is weights plus KV cache plus a 10% activation buffer, calculated from the model's architecture rather than measured on a card.
- Will Mellum2.1 12B-A2.5B Thinking run on a 10GB GPU?
- Yes — at Q4_K_M and 4K context it needs about 8.6 GB, which leaves 1.4 GB spare on a Arc B570 10G, or on any of the other 10 GB cards here (2 in all). That is the smallest size in this index that clears it comfortably; 67 of 78 do.
- Which quantization of Mellum2.1 12B-A2.5B Thinking should I use?
- No published quality comparison exists for this model, so pick by footprint: Q4_K_M is the level most people run, and a higher bits-per-weight level is more faithful. All 2 levels in the index are Q4_K_M, Q8_0. Note this is a mixture-of-experts model — all parameters must be resident even though only a fraction are active per token, so the memory cost follows the total, not the active count.
Where this model fits
Sized at Q4_K_M with a 4K context window, smallest card first. Comfortable means the estimate uses at most 88% of the memory.
How to actually run this
Deployment guides for this model and this class of hardware.
This page's figures change when the model or the runtime does.Last updated 2026-10-13 RSS → /feed.xml