𝗖𝗛𝗜𝗡𝗔'𝗦 𝗖𝗢𝗗𝗜𝗡𝗚 𝗔𝗜 𝗜𝗦 𝗖𝗟𝗢𝗦𝗜𝗡𝗚 𝗧𝗛𝗘 𝗚𝗔𝗣 𝗙𝗔𝗦𝗧
MiniMax released M3. It is an open-weights coding model. It scored 59% on SWE-Bench Pro. It beats GPT-5.5 and Gemini 3.1 Pro.
M3 has a million-token context window. It handles images and video. It costs 5% to 10% of Western models.
Chinese labs are starting a price war. Four labs shipped models in 18 days.
Look at the output costs per million tokens:
- Claude Opus 4.7: $25
- Kimi K2.6: $2.50
- MiniMax M2.7: $1.20
- DeepSeek V4-Flash: $0.28
Performance is catching up. Six months ago, the gap was 20 points. Now, the gap is 10 points.
Western companies sell premium products. Chinese companies make coding AI a commodity.
You want low costs for production. Cheap models with good performance win.
The West builds the most capable AI. China builds the most accessible AI.
The price gap is huge. The performance gap is shrinking.
Source: https://dev.to/max_quimby/chinas-coding-ai-is-closing-the-gap-fast-3o82 Optional learning community: https://t.me/GyaanSetuAi