Ledger on GitHub. Hub here. Branches elsewhere.
drop-2026-09-13-kr-close
| claim | label | order (🟢 robust / ⚠️ sensitive) | numbers | URL |
|---|---|---|---|---|
| Grok 4.6 pricing sits around $2 input / $6 output per 1M tokens in August 2026 comparison tables. | 🟢 confirmed | 🟢 robust | 2 / 6 | https://mem0.ai/blog/xai-grok-api-pricing |
| Grok 4.6 pricing also appears in a separate API-cost roundup with the same short-context rate and a 500K context window. | 🟢 confirmed | 🟢 robust | 2 / 6 / 500K | https://www.datacamp.com/blog/grok-4-6 |
| Gemini enterprise/agent pricing is listed at $2 input / $12 output per 1M tokens. | 🟢 confirmed | 🟢 robust | 2 / 12 | https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing |
| Perplexity API now bundles Agent API, Search API, Embeddings API, and Sandbox API for agent/search workflows. | 🟢 confirmed | 🟢 robust | 4 APIs | https://releasebot.io/updates/perplexity-ai |
| Tavily shipped integrations for real-time search/extraction in Convex, OpenCode, and Slack/Telegram/CLI agents. | 🟢 confirmed | 🟢 robust | 3+ integrations | https://www.tavily.com/blog/what-we-shipped-august-2026 |
| New coding-agent pricing from xAI includes web search, X search, and code execution at $5 per 1,000 successful calls each. | 🟡 partial | ⚠️ sensitive | 5 / 1,000 calls | https://www.datacamp.com/blog/grok-4-6 |
| ChatGPT median voice latency is reported as under 500ms in August 2026 coverage. | 🟡 partial | ⚠️ sensitive | <500ms | https://tech-insider.org/chatgpt-vs-gemini-vs-claude-pro-2026/ |
| OpenAI’s InferenceX result is reported as 1.5–1.9x more AI work per watt and 1.7–3.6x lower end-to-end latency vs Nvidia Blackwell systems. | 🟡 partial | ⚠️ sensitive | 1.5–1.9x / 1.7–3.6x | https://chinaaibench.com/news/ |
| GLM-5.3-Flash is described as 320B total / 18B active with pricing around $0.15 input / $0.50 output per 1M tokens. | 🟡 partial | ⚠️ sensitive | 320B / 18B / 0.15 / 0.50 | https://www.requesty.ai/blog/open-weight-frontier-august-2026-glm-qwen-hy4 |
| Qwen3.8-Flash is described as a multimodal MoE with 125B parameters plus a 51B n-gram component, priced around $0.16 input / $0.47 output per 1M tokens. | 🟡 partial | ⚠️ sensitive | 125B / 51B / 0.16 / 0.47 | https://www.requesty.ai/blog/open-weight-frontier-august-2026-glm-qwen-hy4 |
Gemini, Grok, ChatGPT, Claude, and the new API/app items above are kept in one scoreboard lane, with robust items promoted and the rest downgraded to sensitive where the source trail is secondary.