Pairwise benchmark snapshot · Aug 16, 2026
In this head-to-head showdown, GPT-5.6 Luna delivers higher overall intelligence and human-preferred responses (Elo 1,466 vs 1,320), while GPT-5.6 Luna leads in SWE-bench software engineering benchmarks (61.8%), while Codestral 25.01 is more budget-friendly at $0.9/1M per 1M output tokens, while GPT-5.6 Luna provides faster response streaming (188 tok/s). Review the complete breakdown below to determine which model best fits your performance and budget requirements.
10 direct benchmark disciplines evaluated across capability, speed, and cost
Mistral code-specialist model.
OpenAI 5.6 fast/cheap tier. Official API $0.20/$1.20 per 1M after the Jul 30 80% Luna cut.
Currently comparing Codestral 25.01 vs GPT-5.6 Luna. Alternative family tiers available.
Δ 146 Arena Elo pts
188 tok/s
$0.9 / 1M output
Relative percentile scores computed across all active models in the benchmark catalog.
Comparing top Western standard models with China's leading frontier rival across 6 skill dimensions. Tap any spoke or dot to inspect.
Category wins across reasoning intelligence, generation speed, and token cost.
Preference Elo, Coding proficiency, SWE-bench & LiveBench accuracy
Generation throughput and time to first token responsiveness
Cost per million tokens and max context window length
| Benchmark | Codestral 25.01 | GPT-5.6 Luna | Advantage Delta |
|---|---|---|---|
| Preference Elo | +146 | ||
| Coding Elo | +56 | ||
| LiveBench | +10.2% | ||
| SWE-bench | +10% | ||
| GPQA Diamond | +17% | ||
| Time to first token | +62 ms | ||
| Output speed | 148 tok/s | 188 tok/s | +40 tok/s |
| Input price | $0.3/1M | +$0.1/1M | |
| Output price | $0.9/1M | +$0.3/1M | |
| Context window | +744k |
Simulate monthly production API costs in USD (US Dollar).
Codestral 25.01 is estimated to save $1.00/month ($12/year).
Higher SWE-bench (61.8%).
Lower output list price ($0.9/1M).
Lower TTFT (88 ms).
Larger window (1M).
GPT-5.6 Luna is the side marked multimodal in the catalog.
| Target Workload | Recommended Pick | Evaluation Rationale |
|---|---|---|
| Repo / coding agents | GPT-5.6 Luna | Higher SWE-bench (61.8%). |
| High-volume chat | Codestral 25.01 | Lower output list price ($0.9/1M). |
| Voice / low-latency UI | GPT-5.6 Luna | Lower TTFT (88 ms). |
| Long-document RAG | GPT-5.6 Luna | Larger window (1M). |
| Screenshots / vision | GPT-5.6 Luna | GPT-5.6 Luna is the side marked multimodal in the catalog. |
Luna is the OpenAI fast/cheap 5.6 SKU after Jul 30. It is not Sol. Sort by price and TTFT, not Elo.
High-volume agents that cannot pay Sol. Compare to Luna before you assume the new cheap SKU wins.
Jan 2025 code model. Still a valid cheap-coding compare. Not a chatbot.
No comments posted on this matchup yet. Be the first to share an evaluation note!
Dynamic OpenGraph banner generated at /compare/codestral-25-vs-gpt-5-6-luna/opengraph-image
Verified Head-to-Head Telemetry