MAI-Code-1-Flash Cost Per Task: Cheapest Coding Model?
Microsoft calls MAI-Code-1-Flash its cheapest coding model. In Copilot it bills like Claude Haiku 4.5 (0.33x); its token edge holds only on easy benchmarks.
Chronological archive · page 9 of 10
Independent analysis of the systems, prices and markets behind artificial intelligence.
Archive
Microsoft calls MAI-Code-1-Flash its cheapest coding model. In Copilot it bills like Claude Haiku 4.5 (0.33x); its token edge holds only on easy benchmarks.
Qwen 3.7 Max lists at half Claude Opus 4.8 and runs the same eval suite for a third of the cost. The catch is not hidden cost. It is what the price buys.
Codex leads Claude Code 83.4% to 78.9% on Terminal-Bench 2.1, but the prices match and the cheaper agent flips with your harness. The honest scorecard.
Composer 2.5 finishes a coding task for about $0.07, 10-60x under Claude Opus and GPT-5.5 at near-equal benchmark scores. What the cheap headline leaves out.
A 2026 Microsoft Research preprint found the cheaper-per-token AI model cost more to finish the job in 32% of model pairs. Why the sticker misleads.
An independent replay of 500 Claude Code sessions found rtk, headroom, and caveman cut a $926 bill by just 3.7 percent. Here is why the 60-90% claims miss.
GPT-5.6 launched June 26 2026 as Sol, Terra and Luna. The flagship Sol holds GPT-5.5 pricing at $5/$30 per million tokens. Here is the real cost per task.
Token data from OpenRouter shows open-source LLMs passing proprietary models in mid-2026, a roughly 60/40 flip. The daily breakdown by AI lab.
Open-weights LLMs crossed from toy to useful in 2026. What actually changed, and the cost math for when running a model yourself beats paying an API.