Head-to-head showdown: o3 (Reasoning Frontier) ($10.00 in / $40.00 out per 1M) vs Codestral 2508 (OpenRouter) ($0.30 in / $0.90 out per 1M). Codestral 2508 (OpenRouter) is 41.7× cheaper across standard token mixes, with 1.1M vs 256K context windows.
o3 · openai
codestral-2508 · mistral
Benchmark
| Workload Scenario | o3 (Reasoning Frontier) | Codestral 2508 (OpenRouter) | Price Delta |
|---|---|---|---|
| 1M input tokens (raw text) | $10.00 | $0.30 | +$9.70 |
| 1M output tokens (generation) | $40.00 | $0.90 | +$39.10 |
| 1M tokens · 70% input / 30% output mix | $19.00 | $0.48 | +$18.52 |
| Standard chat turn (4K in / 800 out, 50% cached) | $0.054 | $0.00138 | +$0.0526 |
| Monthly scale (10K requests / day) | $16,200.00 | $414.00 | +$15,786.00 |
Negative difference = o3 (Reasoning Frontier) is cheaper. Positive = Codestral 2508 (OpenRouter) is cheaper.
Scaling Curve
Total cost of a token volume at 70/30 input/output split (uncached). Log-log scale.
Verdict
Codestral 2508 (OpenRouter) 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 (o3 (Reasoning Frontier): 1.1M, Codestral 2508 (OpenRouter): 256K) may justify the premium for your task.
FAQ
On input, Codestral 2508 (OpenRouter) is cheaper ($0.30/M vs $10.00/M). On output, Codestral 2508 (OpenRouter) is cheaper ($0.90/M vs $40.00/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 $0.054 on o3 (Reasoning Frontier) and $0.00138 on Codestral 2508 (OpenRouter) — Codestral 2508 (OpenRouter) is 39.1× more expensive for that workload.
o3 (Reasoning Frontier) supports 1,050,000 tokens (128K max output); Codestral 2508 (OpenRouter) supports 256,000 (204.8K max output). o3 (Reasoning Frontier) fits 4.1× more context, which matters for long documents and agents.