Qwen3.8 Max vs Kimi K3
Qwen3.8 Max and Kimi K3 are the leading Chinese-lab models with opposite strategies. K3 leads the Intelligence Index 60 to 58 and ships open weights; Qwen3.8 Max is closed but costs less hosted — $2/$6 versus $3/$15 per million tokens — and adds video input plus nearly double the serving speed. Self-hosting decides it.
Side by side
| Qwen3.8 Max | Kimi K3 | |
|---|---|---|
| Maker | Alibaba | Moonshot AI |
| Released | 2026-08-03 | 2026-07-16 |
| Context window | 1M tokens | 1M tokens |
| Modalities in | text, image, video | text, image |
| Input $/MTok | $2 | $3 |
| Output $/MTok | $6 | $15 |
| Open weights | No | Yes (Kimi K3 License (commercial terms apply)) |
Benchmark deltas
| Benchmark | Qwen3.8 Max | Kimi K3 | Δ |
|---|---|---|---|
| AA Intelligence Index | 58 | 60 | -2 |
Δ is Qwen3.8 Max minus Kimi K3, in points. Shaded cell wins the row. Dashed values are vendor-published, not yet reproduced — methodology.
Cost calculator
- Qwen3.8 Max
- $160/mo
- Kimi K3
- $300/mo
- Difference
- $140/mo
List prices as of 2026-08-11; excludes caching, batch, and long-context tiers — see model pages for tier notes.
Verdict
Alibaba and Moonshot have swapped the roles they held a year ago: Qwen is now the closed value play and Moonshot the open flagship. K3's two-point Intelligence Index lead and downloadable weights make it the choice wherever sovereignty, fine-tuning, or peak open capability matter. Qwen3.8 Max wins the hosted-economics contest — 60% cheaper on output — while adding video understanding and 77 tokens-per-second serving. API-first products should default to Qwen; infrastructure owners and regulated deployments to K3.
Verdicts are analysis, not measurement, and are labeled as such per our editorial policy.
FAQ
- Is Qwen3.8 Max cheaper than Kimi K3?
- Yes. Qwen3.8 Max costs $2 input / $6 output per million tokens against $3 / $15 for Kimi K3 (prices as of August 11, 2026).
- Which has the longer context window — Qwen3.8 Max or Kimi K3?
- Kimi K3 does — 1M tokens against 1M for Qwen3.8 Max.
- Which is the more capable model, Qwen3.8 Max or Kimi K3?
- Kimi K3 scores higher on the Artificial Analysis Intelligence Index: 60 (measured August 11, 2026) against 58 for Qwen3.8 Max (measured August 11, 2026).
Related news
Comparison published August 11, 2026; underlying data last changed August 11, 2026.