Vals Index: DeepSeek's V4.1 Flash 1017 build is the top open-weight model, 56.58% at $0.64 a test

Vals AI now ranks DeepSeek-V4.1-Flash-1017 first among open-weight models on the Vals Index. It scores 56.58%, #14 of the 50 models on the index as updated October 18. Vals posted the result on X at 05:52 UTC on October 19. It is the first independent board score we have seen for the build DeepSeek shipped on October 17, the open question in our story that day.

THE NUMBERS

56.58% for the 1017 build, up from 51.32% for the September 10 release of V4.1 Flash, which sits at #27. The next open-weight model is Xiaomi's MiMo V2.6 Pro at 55.20% (#15). Directly above DeepSeek is OpenAI's GPT-6 Sol at 57.54%. Gemini 4 Argon leads the whole index at 68.90%.

THE COST

Vals puts the 1017 build at $0.64 per test, against $7.58 for GPT-6 Sol and $0.41 for MiMo V2.6 Pro. Vals lists DeepSeek at $0.30 / $1.20 per million tokens, which is DeepSeek's peak-hour rate; off-peak costs half that, per DeepSeek's pricing page.

THE TOKENS

At the same listed price, the new build cost Vals about twice what the September build did, $0.64 against $0.33 per test, because it used more tokens: 24.72B input and 316.64M output across the index, against 10.05B and 138.82M. DeepSeek's changelog says the 1017 build finishes tasks with shorter output. Vals' run does not show that.

WHAT WE DO NOT KNOW

How Artificial Analysis and the other boards will score the 1017 build, and whether they will list it apart from the September release. The Vals Index is one board, and a rank on it depends on which models Vals has run.