Kimi K3 impressed in a day of API tests, but reasoning tokens made real costs 2–3× Opus
Tarandjpop · reddit · 2026-07-21
A Reddit user spent a day testing Kimi K3 against Claude Opus in production-like image-generation/editing workflows.
Key observations:
- High output quality when Kimi K3 finishes successfully
- Reasoning tokens dominated usage at roughly 70–80% of output tokens
- Some end-to-end requests took close to an hour
- Effective cost for this workload was often 2–3× higher than Opus
- Long-running requests forced infrastructure changes: streaming, larger token budgets, longer-lived connections
- The main reliability issue was peak-hour 429 / capacity errors
The poster says they were impressed by the ceiling, but the model currently has meaningful trade-offs in latency, infrastructure demands, and real-world cost for this workflow. The attached image compares Kimi K3, Opus, and GPT-5.6 Sol in a game-like benchmark-style visual.
Related event: Kimi K3 Test Shows High Performance but 2-3x Cost of Claude Opus(2 posts)→
More from Models
- Anthropic says Opus 5.5 cuts cost per task by ~40% vs Opus 5 — taufiqintech · 2026-10-26
- AI search APIs ranked: top 12 within 7 points, yet prices differ nearly 7x — FinanceYF5 · 2026-10-26
- OpenAI Dots coming to EU/EEA and UK in a few weeks, slides confirm — testingcatalog · 2026-10-26
- OpenAI Dots demoed live as EU launch speculated, Codex adoption in Germany revealed — testingcatalog · 2026-10-26
- "Abliterated models just have brain damage" — community pushes back on unaligned model hype — dhadfieldmenell · 2026-10-26
- dair-ai publishes primer on Cloudflare's Clairic System One extraction model — dair_ai · 2026-10-26