{"slug": "model-recursive-flint-40b-a2-5b", "title": "Model: Recursive Flint 40B-A2.5B", "summary": "Recursive released its open coding model Flint 40B-A2.5B on 21 October under Apache 2.0, with 39.8B total parameters and about 2.5B active per token, built for repository work such as tracing bugs and adding tests. Only 14 of the model's 49 layers keep a cache that grows with context, keeping long sessions cheap: bartowski's Q3_K_M GGUF file is 19.53 GB and needs about 20.6 GB of VRAM at 4K context, while Q4_K_M (25.35 GB) needs about 26.7 GB and so requires a 32 GB card. llama.cpp runs the model as `flint`, and Ollama v0.40.8 supports it on macOS only so far; no per-level quality loss has been published.", "body_md": "# Recursive Flint 40B-A2.5B\n\n40B-A2.5B\nRecursive Flint\n\nRecursive's open coding model, released on 21 October under Apache 2.0: 39.8B parameters, about 2.5B active per token. It is built for repository work such as tracing bugs and adding tests. Only 14 of its 49 layers keep a cache that grows with context, so long sessions stay cheap. On a 24 GB card the Q3_K_M build (bartowski's 19.5 GB file) needs about 20.6 GB at 4K, which is comfortable, and stays only tight up to 128K. Q4_K_M needs about 26.7 GB at 4K, so it wants a 32 GB card; ggml-org's Q4_K_M file is about 2 GB smaller than bartowski's. llama.cpp runs it as `flint`; Ollama v0.40.8 supports it on macOS only so far. Sizes come from real files; no per-level quality loss has been published.\n\n262K\n\nMax Context\n\n3\n\nQuant Variants\n\nGGUF Q8_0\n\nBest Quality\n\n—\n\nPerplexity rise (best build)\n\n## Quantization Variants\n\nPer-quant VRAM, quality loss, and inference speed on RTX 4090\n\nMeasured = run on this site · Estimated = not run here: calculated, or a figure this site has not verified · Community = public reports\n\n| Format | Level | BPW | VRAM | PPL Loss | Speed | Source | Actions | \n|---|---|---|---|---|---|---|---|\n| GGUF | Q3_K_M | 3.92 | 20.6 GB | — | — | Estimated |  | \n| GGUF | Q4_K_M | 5.09 | 26.7 GB | — | — | Estimated |  | \n| GGUF | Q8_0 | 8.51 | 44.5 GB | — | — | Estimated |  | \n\n### Which file each size is for\n\nA file size is not a memory figure: the VRAM column is the calculator's total at 4K context — weights plus KV cache plus a 10% activation buffer. Units differ on purpose: file sizes are decimal GB, as Hugging Face lists them; the VRAM column is GiB (1024³ bytes) with 2% added to the weights, so 7.12 GB of file counts as about 6.8 GiB there. A vision projector (mmproj), where the repo ships one, is a separate download and is in neither number.\n\n- GGUF Q3_K_M[bartowski/RecursiveSI_flint-GGUF](https://huggingface.co/bartowski/RecursiveSI_flint-GGUF) · RecursiveSI_flint-Q3_K_M.gguf · 19.53 GB (19,529,151,712 bytes) · listing read 2026-10-23\n- GGUF Q4_K_M[bartowski/RecursiveSI_flint-GGUF](https://huggingface.co/bartowski/RecursiveSI_flint-GGUF) · RecursiveSI_flint-Q4_K_M.gguf · 25.35 GB (25,346,651,360 bytes) · listing read 2026-10-23\n- GGUF Q8_0[bartowski/RecursiveSI_flint-GGUF](https://huggingface.co/bartowski/RecursiveSI_flint-GGUF) · RecursiveSI_flint-Q8_0.gguf · 42.35 GB (42,350,359,776 bytes) · listing read 2026-10-23\n\n## Similar models\n\n[Compare with Seed-OSS 36B](https://quantized.uk/tools/compare/?a=flint-40b-a2-5b&b=seed-oss-36b)\n\n[36BConsumer GPUPro GPU19.5 GBmin VRAM+1.3%perplexity rise](https://quantized.uk/quant-hub/seed-oss-36b/)\n\n### Seed-OSS 36B Instruct\n\nByteDance Seed\n\nDense 36B with a native 512K context. The Q4_K_M file is 21.8 GB, so with cache and buffers it needs about 23.9 GB at 4K: tight on a 24 GB card, where Q3_K_M (19.5 GB) is the comfortable choice, or a 2×16GB split. The AWQ build is 21.1 GB, not the 4-bit rate the name suggests. Context is the real cost: 512K of KV cache is roughly 128GB on its own, so budget context first and weights second.\n\n[27BConsumer GPUPro GPU17.5 GBmin VRAM—perplexity rise](https://quantized.uk/quant-hub/qwen3-8-27b/)\n\n### Qwen3.8 27B\n\nAlibaba Qwen3.8\n\nHybrid attention: only 16 of its 64 layers keep a KV cache, which is why a 27B model holds a 262K context in about 16GB of cache rather than 64GB. Sizing it as a conventional stack overstates the cache fourfold — the estimates here count the 16 and match published measurements at 8K, 32K and 262K. Quality loss per level has not been published.\n\n[30B-A3BConsumer GPUPro GPU19.2 GBmin VRAM—perplexity rise](https://quantized.uk/quant-hub/glm-4.7-flash/)\n\n### GLM-4.7-Flash\n\nZhipu GLM-4.7\n\nZhipu's lightweight member of GLM-4.7: 30B total, about 3B active per token, MIT licence. Its attention is MLA, which caches one compressed 576-value vector per token per layer instead of full keys and values — so at 32K context the cache is under 2 GB where a conventional 47-layer model would need several times that. llama.cpp runs it as a DeepSeek-2-style model. Sizes below use the calculator's generic Q4_K_M/Q8_0 rates: no GGUF file size could be checked, and quality loss per level has not been published.\n\n[26B-A4BConsumer GPUPro GPU18.2 GBmin VRAM—perplexity rise](https://quantized.uk/quant-hub/gemma-4-26b-a4b/)\n\n### Gemma 4 26B-A4B IT\n\nGoogle Gemma 4\n\nGoogle's Gemma 4 mixture-of-experts model: about 25B parameters in the text model, roughly 4B active per token, a 256K window and image input. Only 5 of its 30 layers keep a cache that grows with context, and those use fewer, wider heads, so the cache stays small — about 0.9 GB at 32K and 5.3 GB at the full 256K. At Q4_K_M that is about 18.8 GB at 32K: comfortable on a 24 GB card, which can still load the whole window (tight, about 23.6 GB). A 16 GB card cannot hold it, since the Q4_K_M file alone is 17.0 GB. The vision encoder is a separate 1.2 GB file on top. Sizes come from bartowski's GGUF files, whose Q4_K_M works out at 5.40 bits per weight, well above the generic 4-bit rate; Unsloth's UD-Q4_K_M is a different file, 0.09 GB smaller. No per-level quality loss has been published.\n\n## Running Recursive Flint 40B-A2.5B locally\n\nAt Q4_K_M and 4K of context, Recursive Flint 40B-A2.5B needs about 26.7 GB — 24.1 GB of weights, 0.19 GB of KV cache and a 2.4 GB activation buffer. The smallest cards in this index that clear that comfortably are 32 GB ones — 4 of them, from the RTX 5090 up — and 24 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.\n\n## What longer context costs\n\nGoing from 4K to 32K adds about 0.77 GB, taking the total to 27.5 GB. That is gentler than the parameter count suggests: only 14 of 49 layers keep a KV cache that grows with context, and 35 more cap theirs at a sliding window, so the cache scales at a fraction of the usual rate. The model's native window is 256K; holding all of it at Q4_K_M would need about 34 GB.\n\n## Which build to download\n\nThis index only tracks Recursive Flint 40B-A2.5B in GGUF, across 3 levels (Q3_K_M, 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.\n\n## Common questions\n\n- How much VRAM does Recursive Flint 40B-A2.5B need?\n- About 26.7 GB at Q4_K_M with 4K of context and batch 1, rising to roughly 27.5 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.\n- Will Recursive Flint 40B-A2.5B run on a 32GB GPU?\n- Yes — at Q4_K_M and 4K context it needs about 26.7 GB, which leaves 5.3 GB spare on a RTX 5090, or on any of the other 32 GB cards here (4 in all). That is the smallest size in this index that clears it comfortably; 24 of 78 do.\n- Which quantization of Recursive Flint 40B-A2.5B should I use?\n- 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 3 levels in the index are Q3_K_M, 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.\n\n## Where this model fits\n\nSized at Q4_K_M with a 4K context window, smallest card first. Comfortable means the estimate uses at most 88% of the memory.\n\n- Runs comfortably on\n- [RTX 509032GB · +5.3](https://quantized.uk/gpu/rtx-5090/)[Arc Pro B70 32G32GB · +5.3](https://quantized.uk/gpu/arc-pro-b70-32g/)[Radeon AI PRO R9700 32G32GB · +5.3](https://quantized.uk/gpu/radeon-ai-pro-r9700-32g/)[Instinct MI100 32G32GB · +5.3](https://quantized.uk/gpu/instinct-mi100-32g/)[A100 40G40GB · +13.3](https://quantized.uk/gpu/a100-40g/)\n- Tight but possible\n- [Mac M3 Pro 36G36GB](https://quantized.uk/gpu/mac-m3-pro-36g/)[Mac M4 Max 36G36GB](https://quantized.uk/gpu/mac-m4-max-36g/)\n- Ships in\n- [GGUF](https://quantized.uk/formats/gguf/)\n- A top pick for\n- [Best local LLM for 24GB](https://quantized.uk/best/24gb-vram/)\n- Try it yourself\n- [Size it in the calculator](https://quantized.uk/tools/vram-calc/?model=flint-40b-a2-5b&quant=Q4_K_M)[Generate the run command](https://quantized.uk/tools/cli-gen/?model=flint-40b-a2-5b&quant=Q4_K_M)\n\n## How to actually run this\n\nDeployment guides for this model and this class of hardware.\n\nThis page's figures change when the model or the runtime does.Last updated 2026-10-23 [RSS → /feed.xml](https://quantized.uk/feed.xml)", "url": "https://wpnews.pro/news/model-recursive-flint-40b-a2-5b", "canonical_source": "https://quantized.uk/quant-hub/flint-40b-a2-5b/", "published_at": "2026-10-23 12:00:00+00:00", "updated_at": "2026-10-23 14:39:50.787261+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-tools", "ai-research"], "entities": ["Recursive", "Flint 40B-A2.5B", "bartowski", "llama.cpp", "Ollama", "ggml-org"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/model-recursive-flint-40b-a2-5b", "markdown": "https://wpnews.pro/news/model-recursive-flint-40b-a2-5b.md", "text": "https://wpnews.pro/news/model-recursive-flint-40b-a2-5b.txt", "jsonld": "https://wpnews.pro/news/model-recursive-flint-40b-a2-5b.jsonld"}}