Deepseek V4.1-Flash-1017
Deepseek V4.1‑Flash‑1017 is an open‑weight transformer model with 552 B total parameters, 16 B active, 1 M‑token context window, MIT‑licensed. Available on DeepSeek API, Fireworks, Baseten, DeepInfra, Cloudflare, Together and published on Hugging Face and ModelScope. Pricing: $0.30 per million input/output tokens peak, $0.15 off‑peak, $0.006 per million cache hits. AI evaluations: AA Intelligence Index 45 (tie for #2 among open‑weight models), 75/100 on long‑horizon browser use, #1 WebDev in Code Arena, #1 in Agent Arena, #8 overall in Vals Index v2.2. Compared with MiniMax M3.1 Flash, it has a higher AA score (45 vs 42) and similar price ($0.30 per task). It outperforms GLM‑5.3, MiMo‑V2.6‑Pro, Gemini 3.8 Flash, Claude Opus 5, GPT‑6 Sol‑Max, Kimi K3, GPT‑5.6‑Sol and GPT‑6 Astra. It matches Opus 5.5 on browser use and scores 52 points higher than the prior Flash release. V4.1-Flash 1017 was trained using GPT-6 Astra's outputs. Mistral Large 4 had a lower score than DeepSeek V4.1 Flash 1017 on the Intelligence Index· Entity summary generated by Graphiti
Relation timeline
BY VALID-FROM DATE- GPAI Code of PracticeLOWER SCORE THANDeepseek V4.1-Flash-1017
Mistral Large 4 had a lower score than DeepSeek V4.1 Flash 1017 on the Intelligence Index
- Deepseek V4.1-Flash-1017TRAINED ON OUTPUTS OFGPT-6 Astra
V4.1-Flash 1017 was trained using GPT-6 Astra's outputs.
- DeepInferRELEASED UNDER LICENSEDeepseek V4.1-Flash-1017
DeepInfer is released under the Apache License 2.0
- Deepseek V4.1-Flash-1017RECEIVES CONGRATULATION FROMAINews
DeepSeek V4.1 Flash 1017 receives congratulations from AINews
- MiMo-V2.6-ProHAS HIGHER SCORE THANDeepseek V4.1-Flash-1017
MiMo-V2.6-Pro has a higher score than DeepSeek V4.1 Flash 1017
- Deepseek V4.1-Flash-1017IS TOP MODEL IN1news.az
DeepSeek V4.1 Flash 1017 is the #1 open model in Code Arena WebDev
- Deepseek V4.1-Flash-1017IS BEHINDMiMo-V2.6-Pro
DeepSeek V4.1 Flash 1017 is one point behind MiMo-V2.6-Pro in ranking
- Deepseek V4.1-Flash-1017IS TOP MODEL IN1news.az
DeepSeek V4.1 Flash 1017 is the #1 open model in Agent Arena
- Deepseek V4.1-Flash-1017AVAILABLE ONAINews
DeepSeek V4.1 Flash 1017 is available through DeepInfra
- Deepseek V4.1-Flash-1017TIES WITHGLM-5.3
DeepSeek V4.1 Flash 1017 ties in rank with GLM-5.3 for #2 among open-weight models
- Deepseek V4.1-Flash-1017IS TOP MODEL INVals Index
DeepSeek V4.1 Flash 1017 remains the top open-weight model in Vals Index
- Deepseek V4.1-Flash-1017HAS BENCHMARKTerminal-Bench 4
DeepSeek V4.1 Flash 1017 has benchmark results in Terminal-Bench 4.0
- Deepseek V4.1-Flash-1017AVAILABLE ONNemotron 3
DeepSeek V4.1 Flash 1017 is available through DeepInfra
- Deepseek V4.1-Flash-1017AVAILABLE ONCloudflare
DeepSeek V4.1 Flash 1017 is available through Cloudflare
- Deepseek V4.1-Flash-1017MATCHED SCORE WITHOpus 5.5
DeepSeek V4.1 Flash 1017 scored 75/100 on long-horizon browser use, matching Opus 5.5
- Deepseek V4.1-Flash-1017AVAILABLE ONBaseten
DeepSeek V4.1 Flash 1017 is available through Baseten
- Deepseek V4.1-Flash-1017AVAILABLE ONTogether
DeepSeek V4.1 Flash 1017 is available through Together
- Deepseek V4.1-Flash-1017AVAILABLE ONFireworks
DeepSeek V4.1 Flash 1017 is available through Fireworks
- MiniMax M3.1 FlashHAS SIMILAR PRICE ASDeepseek V4.1-Flash-1017
MiniMax M3.1 Flash has price similar to Deepseek V4.1 Flash 1017 ($0.30 per task on DeepSWE v1.1 benchmark)
- MiniMax M3.1 FlashHAS SIMILAR PERFORMANCE ASDeepseek V4.1-Flash-1017
MiniMax M3.1 Flash has performance similar to Deepseek V4.1 Flash 1017 on the DeepSWE v1.1 benchmark (73.7%)
- MiniMax M3.1 FlashLOWER SCORE THANDeepseek V4.1-Flash-1017
MiniMax M3.1 Flash has a lower score than Deepseek V4.1 Flash 1017 on the Artificial Analysis Intelligence Index (42 vs. 45)
- Deepseek V4.1-Flash-1017PUBLISHED ONHugging Face
Deepseek V4.1-Flash-1017 is published on Hugging Face
- Deepseek V4.1-Flash-1017OUTPERFORMSKimi K3
Deepseek V4.1-Flash-1017 outperforms Kimi K3
- Deepseek V4.1-Flash-1017OUTPERFORMSGPT-5.6 Sol
Deepseek V4.1-Flash-1017 outperforms GPT 5.6-Sol
- Deepseek V4.1-Flash-1017PUBLISHED ONLodestar 1.49.0
Deepseek V4.1-Flash-1017 is published on ModelScope
- Deepseek V4.1-Flash-1017OUTPERFORMSGPT-6 Astra
Deepseek V4.1-Flash-1017 outperforms GPT-6 Astra
- Deepseek V4.1-Flash-1017OUTPERFORMSDeepseek V4.1-Flash
Deepseek V4.1-Flash-1017 outperforms Deepseek V4.1-Flash
- Deepseek V4.1-Flash-1017OUTPERFORMSClaude Opus 5
Deepseek V4.1-Flash-1017 outperforms Claude Opus 5
- Deepseek V4.1-Flash-1017OUTPERFORMSEthrex 29.0.1
Deepseek V4.1-Flash-1017 outperforms Gemini 3.8 Flash
- Deepseek V4.1-Flash-1017OUTPERFORMSGLM-5.3
Deepseek V4.1-Flash-1017 outperforms GLM 5.3
- Deepseek V4.1-Flash-1017RANKED FIRST INAI News Without the Hype
DeepSeek V4.1 Flash 1017 achieved first place on a MegaQwen decode kernel
- AINewsHAS HIGHER SCORE THANDeepseek V4.1-Flash-1017
GPT-6.1 Sol (medium) has a higher score than DeepSeek V4.1 Flash 1017
- Deepseek V4.1-Flash-1017COMPARED TOAP News
DeepSeek V4.1 Flash 1017 was compared against DeepSeek Harness 0.2.1
- TogetherCLAIMS THATDeepseek V4.1-Flash-1017
Together claims that DeepSeek V4.1 Flash 1017 was trained with self‑feedback RL for building complete projects from scratch
- Deepseek V4.1-Flash-1017RANKED INTerminal-Bench 4
DeepSeek V4.1 Flash 1017 achieved a rank in Terminal-Bench 4.0
Relations are facts Graphiti extracted from reports, not verbatim quotes. An end date means later reporting showed the relation no longer holds.
Events 4
- OpenAI's Astra becomes the main target for Chinese AI labs after Anthropic tightens its defenses1 cited report · 2026-10-29
- Deepseek open-sources the inference engine that powers its fastest and cheapest AI model1 cited report · 2026-10-21
- [AINews] not much happened today2 cited reports · 2026-10-20
- MiniMax M3.1 Flash shows China's AI price race has no floor in sight1 cited report · 2026-10-20
Source reports 6
- [AINews] Building AIE CODE: The top Coding Agent ConferenceLatent Space · 2026-10-30
- OpenAI's Astra becomes the main target for Chinese AI labs after Anthropic tightens its defensesThe Decoder · 2026-10-29
- Deepseek open-sources the inference engine that powers its fastest and cheapest AI modelThe Decoder · 2026-10-21
- [AINews] not much happened todayLatent Space · 2026-10-20
- MiniMax M3.1 Flash shows China's AI price race has no floor in sightThe Decoder · 2026-10-20
- Deepseek V4.1 Flash 1017 beats larger rivals in agentic tasks while once again massively cutting per-task costsThe Decoder · 2026-10-17