The coding specialist duel: Qwen 2.5 Coder 32B ($0.20/$0.60) vs Mistral Codestral 2501 ($0.30/$0.90).
qwen-2.5-coder-32b-instruct · qwen
codestral-2501 · mistral
Benchmark
| Workload Scenario | Qwen 2.5 Coder 32B | Codestral 2501 | Price Delta |
|---|---|---|---|
| 1M input tokens (raw text) | $0.20 | $0.30 | −$0.10 |
| 1M output tokens (generation) | $0.60 | $0.90 | −$0.30 |
| 1M tokens · 70% input / 30% output mix | $0.32 | $0.48 | −$0.16 |
| Standard chat turn (4K in / 800 out, 50% cached) | $0.00092 | $0.00138 | −$0.00046 |
| Monthly scale (10K requests / day) | $276.00 | $414.00 | −$138.00 |
Negative difference = Qwen 2.5 Coder 32B is cheaper. Positive = Codestral 2501 is cheaper.
Scaling Curve
Total cost of a token volume at 70/30 input/output split (uncached). Log-log scale.
Verdict
Qwen 2.5 Coder 32B is cheaper on both input and output rates, so it costs less at every input/output mix. Price alone still isn't the whole decision: capability, latency and context limits (Qwen 2.5 Coder 32B: 128K, Codestral 2501: 256K) may justify the premium for your task.
FAQ
On input, Qwen 2.5 Coder 32B is cheaper ($0.20/M vs $0.30/M). On output, Qwen 2.5 Coder 32B is cheaper ($0.60/M vs $0.90/M). The same model is cheaper on both sides, so it wins at every mix.
A chat-style request (4,000 input + 800 output tokens, 50% cached) costs $0.00092 on Qwen 2.5 Coder 32B and $0.00138 on Codestral 2501 — Qwen 2.5 Coder 32B is 1.5× cheaper for that workload.
Qwen 2.5 Coder 32B supports 128,000 tokens (8.2K max output); Codestral 2501 supports 256,000 (8.2K max output). Codestral 2501 fits 2.0× more context, which matters for long documents and agents.