Know your limits before they know you.

Prices, limits and rankings for 211 AI tools and models: 29 coding tools, 114 LLMs, and 40 video, image and audio models. All linked to the source.

Not sure? Take the 4-question plan picker →

Last verified · 81 of 118 plans verified

211 AI tools and models tracked · 90 official X accounts monitored daily

1 open agent also appears as a coding tool and is counted once.

Most popular 8

More tools by category 21

“From” is the cheapest paid plan in our data, quoted as the source writes it. Chips show every reset window across that tool's plans.

Open agents 11 open-source agents you can run yourself

★ Most talked-about open agent on X this month

Hermes Agent

by Nous Research · MIT

Self-improving personal agent from Nous Research: it writes skills from experience, keeps curated memory, and runs on many backends.

GitHub stars
254,311
Latest release
v0.22.2Oct 15, 2026
Runs
Local + cloud
X posts this month
23of 123 in our brief

Top features

  • Closed learning loopautonomous skill creation, self-improving skills, FTS5 cross-session search
  • Kanban board for coordinating multiple agents
  • Batch trajectory generation for training tool-calling models

Basis: in our X Brief (Sep 16 – Oct 16, 2026), 23 of 123 tallied posts were about Hermes, the most of any agent. OpenClaw was close behind with 22. That's a narrow lead in one month of posts, not a quality ranking. OpenClaw has more GitHub stars.

Compare all 11 open agents →

Latest limit changes

Full changelog →
Show the 5 most recent changes
  1. KimimodelKimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are live in the Kimi API with a 1M-token context; Kimi says full weights will be released by Nov 4, 2026
  2. OpenAI CodexreleaseCodex CLI 0.164.0: /presets, inline arguments for /model and /permissions, Kitty/Sixel image previews in the terminal, enterprise MCP auth, credential masking, and the Bedrock setup wizard on by default
  3. HeyGenHeyGen VideoHeyGen Video API now makes single clips of 5 to 30 seconds for text-to-video and image-to-video, at 480p, 768p, 1080p or 2K; reference-to-video and the MCP stay limited to 15 seconds
  4. Tencent HunyuanmodelTencent open-sourced Hy-Nomas, an 8B interactive video world model: 720p at 24 fps, clips up to 60 s, steered by text prompts and camera paths, with world memory and LoRA fine-tuning; Apache 2.0, runs on a single RTX 4090 (INT8/FP8 with CPU offload)
  5. ReplitintegrationIncorporate your business with Stripe Atlas from inside Replit (Stripe MCP integration), with 20% off when you incorporate through Replit

The weekly diff

What changed in AI coding plans: prices, limits, resets. Email signups open soon. Until then, updates go out on X.

Follow @QuietlyByG on X RSS feed

No email form yet, so we don't collect any email addresses.