Inference Costs for Closed-Source Models Are Now on Par
rwang07 · x · 2026-07-19
A forwarded post compares the gap between Chinese and Western closed-source models regarding inference costs and intelligence per dollar, suggesting they are now roughly on par.
Key points include:
- Opus 4.8 is priced at about $5/$25 per million tokens, while Kimi is around $3/$15, which isn't a massive difference.
- Anthropic's 1P API gross margin is reportedly around 75%, whereas known financials from Chinese labs are closer to 50%.
- This implies that pricing differences likely stem from profit margins rather than inference efficiency gaps.
- The author also notes that training costs and efficiency are harder to gauge, and there's currently insufficient evidence to prove that Chinese labs are significantly more compute-efficient in training.
More from Infra
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11
- LLM Serving Metrics Thread: Why TPOT and Uptime Make or Break User Experience — abhijithneil · 2026-09-11
- PlanetScale launches sharded Postgres: 768 servers acting as one, 1PB scale — dhruv2038 · 2026-09-11