arxiv:2511.15884
Sajjad Pakdamansavoji
savoji
AI & ML interests
VLM, VLA, LLM, Robotics, Computer Vision
Recent Activity
repliedto their post about 4 hours ago
A 284B-parameter DeepSeek-V4-Flash now runs on a 64 GB desktop with 0 GPUs: llama-modchip shrinks the checkpoint to 56.56 GiB (1.71 bpw) and decodes at 9.1 to 14.7 tok/s.
The standard UD-IQ1_S build is 76.87 GiB, above the ~62.5 GiB usable on a 64 GB Linux box. The repo mixes precisions (routed experts mostly IQ1_S, MLA projections in a custom Fractal-64 type) and patches the CPU kernels. The model plus a 256K KV cache then sits at 57.61 to 58.9 GiB RSS with zero swap. The repo was created on 2026-10-29 and already has 744 stars and 470 points on Hacker News.
š The numbers
ā¢ Size: 56.56 GiB, 20.31 GiB smaller than UD-IQ1_S
ā¢ Prefill: 25.0 to 44.36 tok/s (+77%) on a 16-core Ryzen 9 9955HX
ā¢ Decode: 5.91 tok/s unpatched, 9.06 to 9.95 tok/s at Top-6, 11.5 to 14.7 tok/s with the n-gram drafter
ā ļø Caveats
ā¢ "Lossless" means bit-exact against the same re-quantized checkpoint. Versus UD-IQ1_S the author reports 0.9964 top-logit cosine similarity
ā¢ The 98.5/100 quality score comes from the author's own gate, and the headline speeds come from one machine
ā¢ The custom tensor type 40 is rejected by stock llama.cpp, Ollama and LM Studio
š ļø Try it
ā¢ Run scripts/check_cpu_readiness.py from the repo: it needs no model download
ā¢ Run ./modchip bench for speed, then ./modchip eval to check the quality claims yourself
Repo:
https://github.com/synchopate/llama-modchip
HN thread:
https://news.ycombinator.com/item?id=50411961
Would you accept a custom runtime for 9 to 15 tok/s from a 284B model on a 64 GB box?
š Read it elsewhere
ā¢ The deep dive on Medium: https://medium.com/p/e47ab3581f58
ā¢ The long read on Substack: https://sajjadpakdamansavoji.substack.com/p/llama-modchip-runs-a-284b-deepseek
ā¢ The thread on X: https://x.com/saavoji/status/2119858773280141526
ā¢ Every version on sajjadpsavoji.github.io: https://sajjadpsavoji.github.io/blog#llama-modchip repliedto their post about 4 hours ago
A 284B-parameter DeepSeek-V4-Flash now runs on a 64 GB desktop with 0 GPUs: llama-modchip shrinks the checkpoint to 56.56 GiB (1.71 bpw) and decodes at 9.1 to 14.7 tok/s.
The standard UD-IQ1_S build is 76.87 GiB, above the ~62.5 GiB usable on a 64 GB Linux box. The repo mixes precisions (routed experts mostly IQ1_S, MLA projections in a custom Fractal-64 type) and patches the CPU kernels. The model plus a 256K KV cache then sits at 57.61 to 58.9 GiB RSS with zero swap. The repo was created on 2026-10-29 and already has 744 stars and 470 points on Hacker News.
š The numbers
ā¢ Size: 56.56 GiB, 20.31 GiB smaller than UD-IQ1_S
ā¢ Prefill: 25.0 to 44.36 tok/s (+77%) on a 16-core Ryzen 9 9955HX
ā¢ Decode: 5.91 tok/s unpatched, 9.06 to 9.95 tok/s at Top-6, 11.5 to 14.7 tok/s with the n-gram drafter
ā ļø Caveats
ā¢ "Lossless" means bit-exact against the same re-quantized checkpoint. Versus UD-IQ1_S the author reports 0.9964 top-logit cosine similarity
ā¢ The 98.5/100 quality score comes from the author's own gate, and the headline speeds come from one machine
ā¢ The custom tensor type 40 is rejected by stock llama.cpp, Ollama and LM Studio
š ļø Try it
ā¢ Run scripts/check_cpu_readiness.py from the repo: it needs no model download
ā¢ Run ./modchip bench for speed, then ./modchip eval to check the quality claims yourself
Repo:
https://github.com/synchopate/llama-modchip
HN thread:
https://news.ycombinator.com/item?id=50411961
Would you accept a custom runtime for 9 to 15 tok/s from a 284B model on a 64 GB box?
š Read it elsewhere
ā¢ The deep dive on Medium: https://medium.com/p/e47ab3581f58
ā¢ The long read on Substack: https://sajjadpakdamansavoji.substack.com/p/llama-modchip-runs-a-284b-deepseek
ā¢ The thread on X: https://x.com/saavoji/status/2119858773280141526
ā¢ Every version on sajjadpsavoji.github.io: https://sajjadpsavoji.github.io/blog#llama-modchip posted an update about 5 hours ago
Comfy-Org's repackaged MiniMax-H3-Image has 351,803 downloads last month, more than the 142,989 of the official repo, which was created five days earlier.
@MiniMaxAI's 33.1B dense DiT generates about 16 MP directly (4096x4096 up to 6272x2688), with no upscaler, in 6 steps and no classifier-free guidance. One checkpoint covers text-to-image and editing with up to 16 reference images. The card lists recipes for SGLang, vLLM-Omni, diffusers and ComfyUI, and @Comfy-Org added quantized files and workflow templates within days.
š What's new
ā¢ The H3 design without audio layers, 50 layers, with the H3 Qwen3-VL text encoder unchanged
ā¢ Only 6 steps and guidance 1 are accepted; other values return HTTP 400 and negative_prompt is rejected
š ļø Try it
ā¢ Official BF16 DiT is 68.65 GB, plus a separate 66.7 GB text encoder
ā¢ Comfy-Org offers a pruned int8 DiT, int8 or NVFP4 text encoders, an fp16 VAE, and T2I and Edit templates
ā¢ It expects a detailed English prompt: use the MiniMax API or the local MiniMax-H3-Image-PE (17.9 GB, NVFP4)
ā ļø Caveats
ā¢ The card gives no benchmark scores; "strongest instruction following among open-source models" is a vendor claim
ā¢ Community license: written authorization is needed above 1 million USD annualized revenue or for hosted services
https://huggingface.co/MiniMaxAI/MiniMax-H3-Image
https://huggingface.co/Comfy-Org/MiniMax-H3-Image
https://huggingface.co/spaces/MiniMaxAI/minimax-h3-image
Which route would you take for a local run: ComfyUI with the quantized files, or SGLang, vLLM-Omni or diffusers with the official weights?
š Read it elsewhere
ā¢ The deep dive on Medium: https://medium.com/p/f338de809dcd
ā¢ The long read on Substack: https://sajjadpakdamansavoji.substack.com/p/minimax-h3-image-is-a-16-mp-base
ā¢ The thread on X: https://x.com/saavoji/status/2120585989798207554
ā¢ Every version on sajjadpsavoji.github.io: https://sajjadpsavoji.github.io/blog#minimax-h3-imageOrganizations
None yet
