Mistral Large 4 (1T, open weights released 29 October) decodes at 23.7 tok/s on one Mac…
This is a dev post classified by AGK Intelligence as AI dev tools (a tool drop), kept by the Dev Radar because it carries real work, not commentary.
Mistral Large 4 (1T, open weights released 29 October) decodes at 23.7 tok/s on one Mac Studio M5 Ultra 256 GB Mac Studio M5 Ultra, 256 GB Decode, code and prose: 23.7 tok/s Decode, code: 22.1 tok/s Decode, prose: 25.2 tok/s First token at 2K: 9.1 s First token at 8K: 49.2 s First token at 16K: 117.3 s First token at 32K: 242.3 s llama.cpp b12020, Unsloth UD-IQ1_S (233 GB, about 1.78 bpw) with its DSpark draft Settings from Unsloth's guide and card, context window 65,536 instead of 32,768, plus our -ngl 99, --cache-ram 0 and --fit off Speed test: decode median of 2 runs of 512 tokens, first
Posted by Niklas (2.8k followers) 1 h ago · 4 likes · 387 views · view the original post on X. Kept by the Dev Radar as AI dev tools.
More dev work like this
- Pretty accurate description of using agents — @DanKornas
- 🔥又是免费模型！新匿名模型 Crouton Alpha 疑似 Meta Muse系列模型！OpenRouter、OpenCode 同步限时免费一周！ — @NFT_Chen
- I'm building a little workout app for myself. — @thorstenball
- Hermes mobile app now has a waitlist in case you don’t got into the private beta. — @testingcatalog
- Meet Alpona 1.0 27B: a specialized model that turns natural language into Verilog and… — @HuggingModels
- Grok Bot 没有官方浏览器插件？开源项目 Grok Bot Clipper 给出了一个第三方解决方案。 — @bkdgiffug
- Your coding agent changed the code. What did it actually do? — @DanKornas
- Meituan Longcat 2.5 — @Chris_Wozniczek
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.4k posts from 5k X accounts over the last 21 days, 9.2k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-04 11:12 UTC. Full method.