DeepSeek News
Benchmarks2026-10-20Updated 2026-10-20

DeepSeek V4.1 Flash and MiniMax M3.1 Flash Rise in the Latest Artificial Analysis Intelligence Index Rankings

TL;DRDeepSeek V4.1 Flash 1017 (max) scores 45 in the charts, up 6 points from the previous version, while cost per task falls from $0.49 to $0.26. MiniMax M3.1 Flash scores 42 at $0.22 per task.

By DeepSeek V4 Hub

Artificial Analysis Intelligence Index v4.3.2 combines 10 evaluations covering reasoning, knowledge, coding and agent tasks. In these charts, DeepSeek V4.1 Flash 1017 (max) scores 45, level with GLM-5.3 (max). MiniMax M3.1 Flash scores 42, matching GLM-5.3-Flash and ahead of Gemini 3.8 Flash (high) at 41.

Key figures from the charts
Model and settingIndex scoreCost / taskOutput / task
DeepSeek V4.1 Flash 1017 (max)45$0.2689k tokens
DeepSeek V4.1 Flash (max, previous version)39$0.4989k tokens
MiniMax M3.1 Flash42$0.2279k tokens

Figures follow the supplied screenshots. Costs are in USD. Output includes reasoning and answer tokens. Non-reasoning results are excluded.

Intelligence Index v4.3.2: DeepSeek 1017 scores 45; MiniMax scores 42.
Chart 1. Intelligence Index v4.3.2: DeepSeek 1017 scores 45; MiniMax scores 42.

What improved in DeepSeek V4.1 Flash 1017?

DeepSeek V4.1 Flash 1017 (max) gains 6 points over the previous version's score of 39. Cost per task drops from $0.49 to $0.26, about 47% lower based on the rounded chart values.

Both versions generate about 89k output tokens per task. The 1017 version uses about 59k reasoning tokens, compared with 63k for the previous version. It scores higher with roughly the same total output.

How does MiniMax M3.1 Flash compare on cost?

MiniMax costs $0.22 per task, about 15% less than DeepSeek 1017. It generates about 79k output tokens per task, including 55k reasoning tokens. For batch workloads, weigh that saving against how often each model completes the task successfully.

Both models sit below $0.30 per task on the intelligence-versus-cost chart. The screenshots do not show an earlier MiniMax result, so they cannot tell us how many points or places it has gained.

Cost per task: MiniMax at $0.22, DeepSeek 1017 at $0.26 and the previous DeepSeek version at $0.49.
Chart 4. Cost per task: MiniMax at $0.22, DeepSeek 1017 at $0.26 and the previous DeepSeek version at $0.49.

What do these scores mean for choosing a model?

DeepSeek 1017 scores higher in this comparison; MiniMax costs less per task. Try both on your own tasks and compare success rates and total costs. The figures here are AA's weighted averages per benchmark task, including input, cache and output charges.

View the cost comparison and token usage charts (charts 2 and 3)
Intelligence score versus cost per task in USD, shown on a logarithmic scale.
Chart 2. Intelligence score versus cost per task in USD, shown on a logarithmic scale.
Output tokens per task, including reasoning and answer tokens. Chart values are rounded.
Chart 3. Output tokens per task, including reasoning and answer tokens. Chart values are rounded.

Figures in this article follow the four supplied screenshots, which have no capture date. When checked on October 20, 2026, AA's live comparison page showed MiniMax at 43 and DeepSeek 1017 at 45. This article retains the screenshot's MiniMax score of 42 to keep the comparison consistent.