AI Daily · Nov 4, 2026: OpenAI doubles usage limits for Pro and Business Premium tiers
OpenAI doubled all usage limits for Pro 100/200/500 and Business Premium tiers, while Tencent's Hunyuan Hy4 opened via API and Web Chat with 1M-token context and a 50% discount through Dec 31.
Showing the latest briefing (Nov 4, 2026). Every day is generated the next morning.
Pick a date
-
OpenAI doubles usage limits for Pro and Business Premium tiers
OpenAI is doubling all usage limits for the Pro 100/200/500 tiers and Business Premium starting today. Pro 200 limits will not be reduced, so Work and Codex usage stays the same, and extended limits for dots continue for now.
-
Tencent releases Hy4, its first Hunyuan model with native vision
Tencent's Hy4 is out of preview as the first Hunyuan model with native vision and computer use, priced the same as during preview, with open weights and a Hy4 Flash variant promised soon.
-
OpenRouter lists free stealth model Crouton Alpha with 1M context
OpenRouter added stealth reasoning model Crouton Alpha, free to use with a 1M token context and text, image, audio and video input, and no training on user data.
-
OpenCode inference team claims $300 of DeepSeek usage for $10
A member of the OpenCode team said its inference team figured out how to deliver $300 worth of DeepSeek usage for $10. Follow-up posts gave no further technical detail.
-
HeyGen Voice Turbo free on OpenRouter until November 30
HeyGen's lower-latency text-to-speech model Voice Turbo is available free on OpenRouter until November 30, with no credits required and support for 39 languages.
-
Moondream's Parakeet Ultra speech model launches on OpenRouter
Parakeet Ultra, Moondream's post-trained version of NVIDIA Parakeet TDT 0.6B v3, is live on OpenRouter with lower word error rate across benchmarks, 25 European languages, word timestamps, and $0.06 per hour of audio.
-
Command Code raises Ling 3.1 Flash 1019 usage credit to $100
Command Code now gives Ling 3.1 Flash 1019 $100 of usage; the $10 GOAT plan is upgraded to $100 in credits, and all higher plans include GOAT plus their extra usage.
-
RoboColosseum offers 10,000 free robot evaluation hours
Hugging Face and partners launched RoboColosseum, an open robot evaluation initiative with access to 100+ real robots (Unitree G1, LeRobot SO-101, YAM, Franka), 140+ in-the-wild scenes, open rollout data including failures, and 10,000 free evaluation hours.
-
Anthropic introduces FlashProve for provable LLM inference
Anthropic's new FlashProve research uses cryptographic techniques to let developers verify that a response came from a given model, and it is substantially more efficient than prior open-source methods.
-
Hugging Face reproduces weight-editing paper with open-source ML Intern harness
Hugging Face reproduced a paper in which GPT-6 Astra edited Qwen3.5-0.8B weights, changing only 0.016% of them and reportedly improving math performance up to 6x and instruction following 3x; the work uses ML Intern, an open-source harness for AI research and model training.