References 680% CONFIDENCEOverall confidence: 80%How well the pin's source and references back up its dates.Weighted average of how firmly 6 references, the source included, support the pin's start and end times; a reference counts half as much for every 180 days older than the newestShow all pins at 75% confidence or better
The first entry is always the pin's source. Overall confidence is a weighted average of how firmly each reference supports the start and end times used above; a reference counts half as much for every 180 days older than the newest.
- [1]90%huggingface.co/moonshotai/Kimi-K2-Instruct-0905huggingface.co· Posted Oct 12, 2026· Starts Sep 5, 2025 ✓· 23% of score
Sources differ by a few days: the Hugging Face repo's first public discussion is 2025-09-05 (files uploaded 3-4 September), OpenRouter[3] listed it late on 4 September UTC, KrASIA[6] reported on 5 September that Moonshot had "briefly announced" Kimi[2] K2-0905, and Wikipedia's[4] text says 9 September while citing 4 and 5 September sources.
- [2]90%Kimi K2: Open Agentic Intelligence - Kimi Tech Blogkimi.com· Added Oct 12, 2026· 23% of score
Moonshot's own K2 blog carries the banner "Update (0905): We have an updated weight that enhances agentic coding and supports 256K context", linking to the Kimi-K2-Instruct-0905 model page.
- [3]70%Kimi K2 0905 - OpenRouter endpointsopenrouter.ai· Added Oct 12, 2026· 23% of score
OpenRouter's record: created 2025-09-04, 262,144-token context, 1T-parameter MoE with 32B active, listed at $0.60 per million input tokens and $2.50 per million output tokens.
- [4]70%Kimi (chatbot) - Wikipediaen.wikipedia.org· Added Oct 12, 2026· 23% of score
Places Kimi[2]-K2-Instruct-0905 as the September 2025 update that improved coding and raised the context window from 128K to 256K tokens, between the July K2 and the November 2025 K2 Thinking.
- [5]80%Kimi-K2-Instruct-0905 - Simon Willisonsimonwillison.net· Published Sep 6, 2025· 5% of score
6 September hands-on note: a follow-up to the July Kimi[2] K2 with somewhat higher coding scores and context raised from 128,000 to 256,000 tokens, run at about 445 tokens per second on Groq and under a "not-quite-MIT" license.
Suggest a correction
Something missing or wrong? Say it in your own words: a link that backs this pin up, a different start or end date and why, or a fact it lacks or gets wrong. The AI checks it against this pin's sources, searches for better ones, and adds any page that backs you up. The pin's own sources still count most. A picture that shows something else, or shows it badly, is looked at too, and moved down or replaced.