HolyGrAIL

All tasks / Models

DeepSeek-V4.1-Flash-1017

DeepSeek's smaller, lower-cost open-weights Flash MoE line under MIT, multimodal with a 1M-token context.

  • DeepSeek
  • Open weights
  • Released 17 Oct 2026

Official page

Where it ranks

  • A
    Chat & assistants
    Models · Local · 1,475 on LMArena Text Arena (overall, style control)
  • A
    Writing
    Models · Local · 1,440 on LMArena Text Arena: Creative Writing (style control)
  • B
    Coding
    Models · Local · 1,670 on LMArena WebDev / Code Arena (overall)
  • A
    Image understanding
    Models · Local · 1,260 on LMArena Vision Arena (overall, style control)
  • B
    Workflow automation
    Models · Local · 66.6% on AutomationBench-AA (Artificial Analysis)

Facts

Vendor
DeepSeek
Family
DeepSeek Flash
Released
17 Oct 2026
Weights
Open: download and run it yourself
License
MIT
Parameters
552B
GPU memory to run locally
~320 GB (552B backbone parameters per model card (HF safetensors total 763B incl. other modules); 4-bit ≈ 320 GB, estimate.)
API price
$0.3 in / $1.2 out per 1M tokens
Access
Free (open weights)
Last verified
18 Oct 2026

Price from api-docs.deepseek.com, 18 Oct 2026.

Benchmark results

BenchmarkResultEntry on the leaderboardAs ofSource
Artificial Analysis Intelligence Index45DeepSeek V4.1 Flash 1017 (Max)
Raw value from the page's embedded model data; the site displays it rounded to an integer.
19 Oct 2026artificialanalysis.ai
Artificial Analysis MMMU-Pro (independent evaluation)80.6%DeepSeek V4.1 Flash 1017 (Max)
Raw fraction 0.8064 from the model data embedded in https://artificialanalysis.ai/leaderboards/models, converted to percent; rank 24 of 124 entries with an MMMU-Pro result; no last-updated date shown.
19 Oct 2026artificialanalysis.ai
AutomationBench-AA (Artificial Analysis)66.6%DeepSeek V4.1 Flash 1017 (Max)
AutomationBench-AA Score (share of each task's objectives completed, a task with any guardrail violation scoring zero); raw value from the page's embedded model data (displayed rounded); rank 12 of 32; Tasks Completed 36.68%, Objectives Completed 81.33%; no last-updated date shown, so as_of = retrieval date. The earlier checkpoint 'DeepSeek V4.1 Flash (Max)' scores 68.89; only the current 1017 checkpoint is recorded.
19 Oct 2026artificialanalysis.ai
EQ-Bench Creative Writing v31,540deepseek-v4.1-flash
Elo score from the leaderboard data embedded in https://eqbench.com/creative_writing.js (file last modified 2026-09-24); rank 58 of 140; rubric score 16.08. Marked as a new entry on the board. Board predates the DeepSeek-V4.1-Flash-1017 snapshot (released 2026-10-17); mapped to the current record the same way as its existing lmarena-text score.
24 Sep 2026eqbench.com
LMArena Text Arena (overall, style control)1,475deepseek-v4.1-flash-max
From LMArena's official leaderboard-dataset (text_style_control, split latest, category overall, published 2026-10-15), which mirrors https://arena.ai/leaderboard/text (site blocks automated fetches). Score is for the 2026-09-10 DeepSeek-V4.1-Flash snapshot (leaderboard published before the 1017 update).
15 Oct 2026huggingface.co
LMArena Text Arena: Creative Writing (style control)1,440deepseek-v4.1-flash-max
From LMArena's official leaderboard-dataset (text_style_control, split latest, category creative_writing, published 2026-10-15), which mirrors https://arena.ai/leaderboard/text/creative-writing (site blocks automated fetches). Rank 58 of 414; 2597 votes; 95% CI 1427.6-1452.4. Board predates the DeepSeek-V4.1-Flash-1017 snapshot (released 2026-10-17); mapped to the current record the same way as its existing lmarena-text score.
15 Oct 2026huggingface.co
LMArena Vision Arena (overall, style control)1,260deepseek-v4.1-flash-max
From LMArena's official leaderboard-dataset (vision_style_control, split latest, category overall, published 2026-10-09), which mirrors https://arena.ai/leaderboard/vision (site blocks automated fetches). Rank 46 of 165; 3559 votes; 95% CI 1249.4-1271.2. Board predates the DeepSeek-V4.1-Flash-1017 snapshot (released 2026-10-17); mapped to the current record the same way as its existing lmarena-text score.
9 Oct 2026huggingface.co
LMArena WebDev / Code Arena (overall)1,670DeepSeek-V4.1-Flash-20261017 (Max)
From LMArena's official leaderboard-dataset (webdev = Code Arena, split latest, category overall, published 2026-10-18), which mirrors https://arena.ai/leaderboard/webdev. Rank 12; 1417 votes; 95% CI 1653.5-1687.4.
18 Oct 2026huggingface.co

Where you can use it

Earlier versions

  • DeepSeek-V4.1-Flash · 10 Sep 2026
  • DeepSeek-V4-Flash-0731 · 31 Jul 2026
  • DeepSeek-V4-Flash · 24 Apr 2026

Compare

Alternatives to DeepSeek-V4.1-Flash-1017

Sources