Ledger on GitHub. Hub here. Branches elsewhere.
drop-2026-09-09-kr-close
| claim | label | order | numbers | URL |
|---|---|---|---|---|
| Grok 4.6 pricing stayed at $2/M input and $6/M output; Cursor also listed a cached-input rate of $0.50/M. | 🟢 robust | 1 | 2 / 0.50 / 6 | https://cursor.com/grok |
| Grok 4.6 docs add a Fast tier at $4/M input, $1/M cached input, and $12/M output. | 🟢 robust | 2 | 4 / 1 / 12 | https://cursor.com/docs/models/grok-4-6 |
| Grok 4.6 was positioned for coding and agent workflows, with launch distribution in Cursor and Grok Build. | ⚠️ sensitive | 3 | — | https://www.developersdigest.tech/blog/weekly-highlights-2026-08-14 |
| Qwen3.8-2.4T-A95B was described as an open-weight MoE with about 2.4T total parameters and about 95B active. | 🟢 robust | 4 | 2.4T / 95B | https://ai-solutions.wiki/news/open-weight-models-august-2026/ |
| Qwen3.8 Flash pricing was reported around $0.14–$0.16/M input and $0.42–$0.47/M output. | ⚠️ sensitive | 5 | 0.14–0.16 / 0.42–0.47 | https://developer.puter.com/tutorials/qwen-api-pricing/ |
| GLM-5.3 weights were posted after the API release, with Hugging Face timing reported in late August. | ⚠️ sensitive | 6 | late Aug | https://ai-solutions.wiki/news/open-weight-models-august-2026/ |
| GLM-5.3 was described as 753B total parameters in secondary coverage. | ⚠️ sensitive | 7 | 753B | https://truescho.com/en/blog/glm-5-3-open-weights-release-2026 |
| GLM-5.3-Flash was reported as 320B total, 18B active, with 1M context and MIT licensing. | ⚠️ sensitive | 8 | 320B / 18B / 1M | https://www.requesty.ai/blog/open-weight-frontier-august-2026-glm-qwen-hy4 |
| Tencent Hy4 was reported as a 770B open model with 49B active parameters and 1M context. | ⚠️ sensitive | 9 | 770B / 49B / 1M | https://www.forbes.com/sites/jonmarkman/2026/08/31/tencent-open-sources-hy4-its-770-billion-parameter-flagship-model/ |
| Gemini 3.1 Pro was listed at $2/M input and $12/M output, with higher rates above 200K tokens. | ⚠️ sensitive | 10 | 2 / 12 | https://codersera.com/blog/grok-4-6-pricing-api-costs-2026/ |
| Gemini 2.5 Flash-Lite was listed at 0.28s time-to-first-token in one August latency ranking. | ⚠️ sensitive | 11 | 0.28s | https://modelrefs.com/news/state-of-the-api-2026-08/ |
| Claude, ChatGPT, Gemini, and Grok were compared in a broad pricing table, but vendor-doc confirmation was not present in the gathered set. | ⚠️ sensitive | 12 | n/a | https://aitoolsreview.co.uk/insights/claude-vs-chatgpt-gemini-grok |