Kimi K2.6 Tops Vals Index Among Open-Weight Models
Benchmark results published by Vals AI rank Kimi K2.6 as the top open-weight model on the Vals Index and No. 8 overall. The model also led open-weight peers on Terminal Bench 2 and SWE-Bench, improving 17 points and 8 points over K2.5, respectively, while placing fourth overall on SWE-Bench and fifth on LiveCodeBench.
Vals AI said K2.6 costs $0.21 per test on average, matching GLM 5.1 and running about five times cheaper than Opus 4.7 at $1.05 per test. Vals AI said the model is competitive with many closed-weight models at a fraction of their price.
From the sources (6 posts)
@valsaiCoding is where K2.6 excels. It took #1 among open-weight models on Terminal Bench 2 (+17 points over K2.5) and SWE-Bench (+8 points), and broke into the top 5 overall on both SWE-Bench (#4) and LiveCodeBench (#5), rivaling frontier closed-
@valsaiK2.6 costs $0.21 per test on average, the same as GLM 5.1 and roughly 5x cheaper than Opus 4.7 at $1.05. Moonshot raised prices from K2.5 ($0.10/$3 → $0.16/$4), but K2.6 still is a far cheaper model than the frontier closed source models.
@valsaiResults on the remainder of our benchmarks are in for Kimi K2.6, the new #1 open-weight model on the Vals Index (#8 overall) Kimi K2.6 is competitive with many closed-weight models, at a fraction of their price.
@valsaiCoding is where K2.6 excels. It took #1 among open-weight models on Terminal Bench 2 (+17 points over K2.5) and SWE-Bench (+8 points), and broke into the top 5 overall on both SWE-Bench (#4) and LiveCodeBench (#5), rivaling frontier closed-
@teortaxestexRT @ValsAI: Results on the remainder of our benchmarks are in for Kimi K2.6, the new #1 open-weight model on the Vals Index (#8 overall) K…
@scaling01RT @ValsAI: Results on the remainder of our benchmarks are in for Kimi K2.6, the new #1 open-weight model on the Vals Index (#8 overall) K…