Good Morning!
Nothing entered or left the board overnight, but it swelled by about a third and the order reshuffled. Hy4 preview climbed from sixth to third, and GPT-5.6 Luna barged in behind it. MiniMax M3.1 Flash was shoved down five places, with DeepSeek V4 Flash slipping after it.
Hy4 pulled in a flood of new requests at the same heavy size, and GPT-5.6 Luna drew more people into general work and gave each longer jobs. MiniMax and DeepSeek V4 Flash kept the same callers sending the same work. On a rising board, standing still looks like falling.
Model | tokens · vs. previous day (Nov 1) | ||
1 | DeepSeek V4.1 Flash · DeepSeek | 10.20T | – +25% |
2 | GLM 5.3 Flash · Z.ai | 2.59T | – +83% |
3 | Hy4 preview · Tencent | 1.58T | ▲3 +115% |
4 | GPT-5.6 Luna · OpenAI | 1.32T | ▲4 +130% |
5 | GPT-6 Luna · OpenAI | 1.13T | – +39% |
6 | Claude Opus 5.5 · Anthropic | 0.98T | ▲1 +49% |
7 | DeepSeek V4 Flash · DeepSeek | 0.93T | ▼3 +13% |
8 | MiniMax M3.1 Flash · MiniMax | 0.91T | ▼5 +1% |
9 | Kimi K3.1 Pro · MoonshotAI | 0.72T | – +26% |
10 | GPT-6.1 Sol · OpenAI | 0.55T | – +30% |
Deep Principle Opens Hydrogen-1.0 Weights

TL;DR: Deep Principle released Hydrogen-1.0, an open-weight scientific research model post-trained from GLM-5.3 on internal research data, lab records and discovery workflows. The team reports 38.3% on FrontierScience Research versus 33.3% for GPT-6 Astra, and 93.0% on GPQA Diamond. The weights are available on Hugging Face. Read more →
Anthropic Raises Its Google Tab

TL;DR: Anthropic’s amended IPO filing puts its minimum Google Cloud spend at $121.3 billion through July 2033, up $6.8 billion from its initial filing. It separately estimates $125 billion to $150 billion for additional TPU capacity expected from 2028 through 2031, with payments depending on capacity, delivery and financing. Read more →
Grok Bot Users Hit Tighter Limits

TL;DR: Some Grok Bot users reported that their weekly allowances suddenly maxed out over the weekend, including one who went from 50% to 100%. Elon Musk said the cost of serving the agent was unsustainably high relative to revenue. Read more →
Claude Code Steers Away From Game Remakes

TL;DR: Anthropic adjusted the Opus 5.5 prompt in Claude Code to steer it away from recreating specific existing games. Engineer Addy Osmani said the prompt still allows original games, games in established genres and projects using the developer’s own designs and assets. Read more →
Terra Costs Less Across Three Tests

TL;DR: In three synthetic tasks run at maximum effort by The New Stack, GPT-6.1 Terra cost less and finished faster than Claude Opus 5.5. It tied Opus on release triage, scored higher on a cell-lineage task and completed a CI audit that Opus did not finish. Read more →

Top AI models. Lower costs.
Access GPT, Claude, Gemini, and more through one OpenAI-compatible API—and save up to 70% on AI API costs. Create your DIT account, get your API key, and start building.