Side-by-side comparison of pricing, 12 benchmarks, and generation speed.
| Metric | Qwen3 235B A22B (Non-reasoning) | Kimi K2.5 (Reasoning) |
|---|---|---|
| Input ($/M tokens) | $0.7 | $0.6 |
| Output ($/M tokens) | $2.8 | $2.75 |
Verdict. Kimi K2.5 (Reasoning) takes the aggregate benchmark matchup 2–0 across 2 categories. Real workloads usually care about a handful of specific tasks — see the per-benchmark table above.
Pricing. Both models sit in the budget bracket for output-token pricing. At 1.0× the per-million-token cost, Kimi K2.5 (Reasoning) is meaningfully cheaper if your traffic is output-heavy (long completions, document generation, agent loops). Kimi K2.5 (Reasoning) makes more sense when output volume is low and absolute reasoning quality justifies the premium.
Strengths. Qwen3 235B A22B (Non-reasoning) is strongest on Intelligence Index (8.3). Kimi K2.5 (Reasoning) leads on Coding Index (46.8), Intelligence Index (23.5).
Speed. Speed data is incomplete for this pair; benchmark and price should decide.
Provider. Alibaba and Kimi sell to overlapping but distinct developer audiences: Alibaba tends to ship frontier reasoning models with premium positioning, while Kimi often prices more aggressively. Your existing vendor relationships, billing, and SLA preferences may matter as much as the raw numbers above.
Workload cost. Workload scenarios (per million requests at 30M input + 15M output tokens): Qwen3 235B A22B (Non-reasoning) costs $63.00 ($756/year); Kimi K2.5 (Reasoning) costs $59.25 ($711/year). At a smaller 5M-input/2M-output scale (single-developer tool or prototype): Qwen3 235B A22B (Non-reasoning) ≈ $9.10/run, Kimi K2.5 (Reasoning) ≈ $8.50/run. At agent/realtime scale (200M input / 100M output per million requests): Qwen3 235B A22B (Non-reasoning) ≈ $420/run, Kimi K2.5 (Reasoning) ≈ $395/run. Kimi K2.5 (Reasoning) becomes more attractive at higher volume — the absolute per-token pricing difference compounds when you ship at scale.
Recommendation. Both models have legitimate use cases — the right answer depends on whether you are optimizing for benchmark ceiling, latency, or unit cost. Start with the cheaper / faster model, evaluate against your specific task, and only switch if the upgrade shows a meaningful lift.
Data from Artificial Analysis API — 12 benchmarks
Kimi K2.5 (Reasoning) is cheaper overall. Its blended price (3:1 input/output ratio) is $1.14/M tokens vs $1.23/M for Qwen3 235B A22B (Non-reasoning).
Kimi K2.5 (Reasoning) wins 2 out of 12 benchmarks compared to 0 for Qwen3 235B A22B (Non-reasoning). See the detailed benchmark chart above for per-category results.
Qwen3 235B A22B (Non-reasoning) generates tokens faster at 60 tok/s vs — tok/s. Qwen3 235B A22B (Non-reasoning) also has lower time-to-first-token (2.74s vs —s).
Choose based on your priorities: Kimi K2.5 (Reasoning) for lower cost, Kimi K2.5 (Reasoning) for stronger benchmark performance, and Qwen3 235B A22B (Non-reasoning) for faster generation. For latency-sensitive apps, check the TTFT comparison above.