Compare / head-to-head
Kimi K2vs
Qwen3 235B-A22B Instruct 2507
Qwen3 235B-A22B Instruct 2507 leads 3 of 5 shared benchmarks. Qwen3 235B-A22B Instruct 2507 has the larger context window (262K tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
2 – 3
Qwen3 235B-A22B Instruct 2507 leads
Cheaper per token
—
list price, input + output
Larger context
Qwen3 235B-A22B Instruct 2507
262K tokens
Providers
2 providers
Moonshot AI · Qwen
Kimi K2
Moonshot AI · released 2025-07-11
text
- Context
- 128K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 18 · 7 core
Qwen3 235B-A22B Instruct 2507
Qwen · released 2025-07-21
text
Quality
Benchmark matrix
5 shared · 13 only Kimi K2 · 12 only Qwen3 235B-A22B Instruct 2507| Benchmark | Kimi K2 | Qwen3 235B-A22B Instruct 2507 | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 5 | ||||
| MMLU-Pro | 81.1% ↗ | 83% ↗ | -1.9 pt | Qwen3 235B-A22B Instruct 2507 |
| Aider-Polyglot | 60% ↗ | 57.3% ↗ | +2.7 pt | Kimi K2 |
| MMLU-Redux | 92.7% ↗ | 93.1% ↗ | -0.4 pt | Qwen3 235B-A22B Instruct 2507 |
| SimpleQA | 31% ↗ | 54.3% ↗ | -23.3 pt | Qwen3 235B-A22B Instruct 2507 |
| WeirdML (Håvard Tveit Ihle) | 39.4% ↗ | 38.7% ↗ | +0.7 pt | Kimi K2 |
| Only Kimi K2 reports · 13 | ||||
| AIME 2025 | 49.5% ↗ | not reported | — | — |
| GPQA-Diamond | 75.1% ↗ | not reported | — | — |
| GSO Opt@1 (GSO) | 2% ↗ | not reported | — | — |
| Humanity's Last Exam (Text Only) | 4.7% ↗ | not reported | — | — |
| IFEval (Prompt Strict) | 89.8% ↗ | not reported | — | — |
| LiveCodeBench v6 (Aug 24 - May 25) | 53.7% ↗ | not reported | — | — |
| SimpleBench (SimpleBench) | 26.3% ↗ | not reported | — | — |
| SWE-bench Multilingual (Agentic Coding, single attempt) | 47.3% ↗ | not reported | — | — |
| SWE-bench Verified (Agentic Coding, single attempt) | 65.8% ↗ | not reported | — | — |
| Tau2 airline | 56.5% ↗ | not reported | — | — |
| Tau2 retail | 70.6% ↗ | not reported | — | — |
| Tau2 telecom | 65.8% ↗ | not reported | — | — |
| TerminalBench (Inhouse Framework) | 30% ↗ | not reported | — | — |
| Only Qwen3 235B-A22B Instruct 2507 reports · 12 | ||||
| LMArena Elo | not reported | 1422.4 ↗ | — | — |
| AIME25 | not reported | 70.3% ↗ | — | — |
| ARC-AGI | not reported | 41.8% ↗ | — | — |
| Arena-Hard v2 | not reported | 79.2% ↗ | — | — |
| BFCL-v3 | not reported | 70.9% ↗ | — | — |
| GPQA | not reported | 77.5% ↗ | — | — |
| HMMT25 | not reported | 55.4% ↗ | — | — |
| IFEval | not reported | 88.7% ↗ | — | — |
| LiveBench 20241125 | not reported | 75.4% ↗ | — | — |
| LiveCodeBench v6 (25.02-25.05) | not reported | 51.8% ↗ | — | — |
| SuperGPQA | not reported | 62.6% ↗ | — | — |
| TAU2-Retail | not reported | 74.6% ↗ | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Kimi K2 minus Qwen3 235B-A22B Instruct 2507 in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | Kimi K2 | Qwen3 235B-A22B Instruct 2507 | Edge |
|---|---|---|---|
| Context window | 128K tokens | 262K tokens | Qwen3 235B-A22B Instruct 2507 |
| Max output | — | — | — |
| Input price / 1M | — | $0.23 ↗ | — |
| Output price / 1M | — | $0.92 ↗ | — |
| Cached input / 1M | — | — | — |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | — | $1.15 | — |
| Modalities | text | text | Tie |
| Released | 2025-07-11 | 2025-07-21 | — |
| Cited benchmark scores | 18 | 17 | — |
Reliability
Provider status
Live from /statusMore matchups
Kimi K2 vs …
Models sharing the most benchmarksMore matchups
Qwen3 235B-A22B Instruct 2507 vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Kimi K2 and Qwen3 235B-A22B Instruct 2507 on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.