DeepSeek V4.1 Flash non-reasoning hits ~1,486 tokens/s on Artificial Analysis at Intelligence Index 18.5

teortaxesTex · x · 2026-10-31

• Artificial Analysis added a non-reasoning entry for DeepSeek V4.1 Flash: Intelligence Index 18.5 with a median output speed of 1,486 tokens/s (the author measured 1,488 t/s and 1.5s end-to-end latency). The Max (reasoning) variant scores 32.5 at 175 t/s.
• The gap is stark: on FrontierCode 2.0 Max hits 28.5% vs 6.1% non-reasoning, though Omniscience flips (2.1 vs 0.9) and AA-Briefcase is close (1570 vs 1426).
• Notably, the fast variant isn't cheaper: $0.45/$1.80 per 1M tokens vs $0.30/$1.20 for Max. The takeaway: tool-heavy, short-turn loops benefit from the extreme speed, while reasoning-heavy tasks still need Max.
• Benchmarks shown also put the non-reasoning variant on par with MiMo 2.6-Pro and Mistral 4 Large on some tasks.

Original post →

More from Models

• Leaker Account Teases Major Unannounced OpenAI Drop From Sam Altman — iruletheworldmo · 2026-10-31
• Blogger argues $30K local Kimi K3.1 setup is worth it, despite taking 4.76 years to match GPT-6.1 Terra output — bookwormengr · 2026-10-31
• SGLang boosts DeepSeek V4.1 Flash throughput 67% to 1.09M tok/s per GPU in 4 days — BanghuaZ · 2026-10-31
• DeepSeek ranks near-zero on AA-Omniscience hallucination test, sparking leaderboard credibility fight — teortaxesTex · 2026-10-31
• Dev tasks Claude Fable 5.1, GPT Astra and Grok 4 with overnight long-running job comparison — therealdanvega · 2026-10-31
• Qwen4 with HC architecture could widen the Q4 vs Q6 quantization gap significantly — Ok-Shower7286 · 2026-10-31

Models channel →