Kimi K2.5+ Fingerprint Spotted, Seen as GLM 5-Class Base with Cheaper Inference
teortaxesTex · x · 2026-10-17
A new fingerprint for Kimi-for-coding (K-2.8) was spotted (fpv0a3373eb3 → fpv0404e18bd), hinting Kimi K2.5+ is near. teortaxesTex argues:
- K2.5+ is in principle about as good a base as GLM 5
- With MLA, Moonshot could follow Zhipu in adopting DSA+ sparse attention to slash inference costs, though still behind V4.1/V4
- It could also distill from K3.1; the author still calls the strategy "weird"
More from Models
- Full FP8 n-gram table visibly improves local Qwen3.8-Flash-Next at zero speed cost — sillynoobhorse · 2026-10-17
- Altman says OpenAI shelved a ready-to-launch model as too risky, warning of AI 'escape' scenarios — Aiden_Tech_Ai · 2026-10-17
- Audit of 160,860 citations: nearly 40% of links shown by Chinese AI search go unused — Sanbi_Ai · 2026-10-17
- Claude Trained on the Entire Web Now Refuses to Translate Web Pages — lxfater · 2026-10-17
- Common Sense Media rates ChatGPT for Teens 'Unacceptable Risk' as 8 states sue OpenAI — jurgo123 · 2026-10-17
- Local Nemotron 3.5 on one A100 hits 169 tok/s, processes 100 clinical notes for 2.4 cents — MaziyarPanahi · 2026-10-17