Kimi K3 is claimed to have 2.8 trillion parameters and a sharp token-price gap
thisguyknowsai · x · 2026-07-22
A repost claims Kimi K3 was released last week with 2.8 trillion parameters and weights landing on July 27.
The post says it benchmarks just behind Claude Fable 5 and GPT-5.6, while also highlighting a price comparison for output tokens:
- Claude Fable 5: $50 per million output tokens
- Kimi K3: $15
- GLM-5.2: $4.40
- DeepSeek V4: $0.87
The argument is that for repetitive document-extraction workloads, the quality gap between frontier models and good open models is shrinking, but the price gap remains large. The poster also notes that teams often keep using the pipeline model they already chose rather than re-testing newer options.
Related event: Moonshot Releases 2.8 Trillion Parameter Open-Weight Model Kimi K3(8 posts)→
More from Models
- Benchmark scores drop from 89% to 19% on new evals — how benchmaxxing breaks leaderboard trust — airesearch12 · 2026-09-11
- Developer Building a Unified Leaderboard of All Model Benchmark Scores — airesearch12 · 2026-09-11
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11
- GPT-5.6 writes well but is instantly forgettable, user complains — BasedRaddka · 2026-09-11
- Opus Refuses Protein Research Codebase Over 'Safety' Concerns, Dev Considers Rolling His Own — josephdviviano · 2026-09-11
- User Hails Unconfirmed 'DeepSeek 4.1 Flash' as an Inflection Point in LLMs — himanshustwts · 2026-09-11