# Agent Stack Index (updated 2026-10-28)

Machine-readable: https://gofreealts.com/agent-stack/data.json. Each row: rank. name: value (unit), source url.

## Tools agents call most
Unit: calls, 7 days. Source: https://monid.ai/rankings. Checked: 2026-10-28.

Note: Calls routed through Monid, data through 2026-10-27. Free calls count 1/5. Only Monid traffic.

1. TinyFish: 1.42M (top endpoint: search) <https://monid.ai/rankings>
2. TikHub: 1.15M (top endpoint: tiktok/app/v3/fetch_video_comments) <https://monid.ai/rankings>
3. AnyAPI: 945.2K (top endpoint: linkedin.company) <https://monid.ai/rankings>
4. Litescrape: 472.4K (top endpoint: google/maps) <https://monid.ai/rankings>
5. Firecrawl: 437.8K (top endpoint: search) <https://monid.ai/rankings>
6. Context.dev: 423.9K (top endpoint: web/scrape/markdown) <https://monid.ai/rankings>
7. Hunter: 330.8K (top endpoint: domain-finder) <https://monid.ai/rankings>
8. DataForSEO: 295.0K (top endpoint: serp/google-maps) <https://monid.ai/rankings>
9. Apify: 275.7K (top endpoint: apidojo/tiktok-profile-scraper) <https://monid.ai/rankings>
10. MrScraper: 247.8K (top endpoint: serp/google) <https://monid.ai/rankings>
11. ContactOut: 226.8K (top endpoint: people/count) <https://monid.ai/rankings>
12. Apollo: 157.9K (top endpoint: people/match) <https://monid.ai/rankings>
13. DefiLlama: 63.7K (top endpoint: coins/prices/historical/{timestamp}/{coins}) <https://monid.ai/rankings>
14. Ploid: 61.9K (top endpoint: linkedin/profile) <https://monid.ai/rankings>
15. GPT Images: 60.9K (top endpoint: image/gpt-image-2.5-sunburst-edit) <https://monid.ai/rankings>
16. Exa: 49.4K (top endpoint: search) <https://monid.ai/rankings>
17. Simple FS: 39.7K (top endpoint: rm) <https://monid.ai/rankings>
18. TypeSafe: 34.6K (top endpoint: systemone) <https://monid.ai/rankings>
19. SerpApi: 28.2K (top endpoint: amazon/search) <https://monid.ai/rankings>
20. Search1API: 27.9K (top endpoint: search) <https://monid.ai/rankings>

## Most called endpoints
Unit: calls, 7 days. Source: https://monid.ai/rankings. Checked: 2026-10-28.

1. TinyFish Web Search: 981.1K (Web search · 342 agents) <https://monid.ai/rankings>
2. TinyFish Fetch: 354.5K (Web search · 289 agents) <https://monid.ai/rankings>
3. Firecrawl Search: 347.7K (Web search · 32 agents) <https://monid.ai/rankings>
4. Scrape Markdown: 343.3K (Web scraping & extraction · 287 agents) <https://monid.ai/rankings>
5. Search Google Maps: 269.1K (Maps, reviews & local · 104 agents) <https://monid.ai/rankings>
6. Search Google Results: 235.6K (Web search · 136 agents) <https://monid.ai/rankings>
7. Count Matching Profiles: 210.0K (Lead gen · 34 agents) <https://monid.ai/rankings>
8. Find Company Domain: 139.8K (Lead gen · 50 agents) <https://monid.ai/rankings>
9. LinkedIn Company: 139.5K (Social · 23 agents) <https://monid.ai/rankings>
10. TikTok Video Comments: 113.4K (Social · 25 agents) <https://monid.ai/rankings>

## Agent harnesses by usage
Unit: usage score 0-100. Source: https://github.com/ (stars) · api.npmjs.org · openrouter.ai/apps · agentgid.com. Checked: 2026-10-28.

Note: Score = 35% OpenRouter app tokens + 35% npm/PyPI weekly downloads + 15% GitHub stars + 15% AgentGid score (log-scaled). OpenRouter only sees BYO-key traffic.

1. Claude Code: 91.6 (8,641,472 npm/wk · 153,684★ · v2.1.312) <https://claude.com/product/claude-code>
2. OpenAI Codex CLI: 85.1 (18,127,935 npm/wk · 131,394★ · rust-v0.165.0) <https://developers.openai.com/codex>
3. Pi: 83.5 (4,099,842 npm/wk · 118,471★ · v1.4.0) <https://github.com/earendil-works/pi>
4. OpenClaw: 79.7 (2,722,524 npm/wk · 393,255★ · v2026.10.8) <https://openclaw.ai>
5. Hermes Agent: 72.0 (257,948★ · v0.22.5) <https://hermes-agent.nousresearch.com/>
6. DeepSeek Harness: 71.8 (438,180 npm/wk · 264,493★ · dsh-v0.2.2-alpha.2) <https://deepseek.com/harness>
7. Oh My Pi (omp): 62.9 (259,789 npm/wk · 36,928★ · v18.13.2) <https://omp.sh>
8. Kilo Code: 58.1 (37,095 npm/wk · 27,775★ · v7.8.18) <https://kilocode.ai>
9. Cline: 56.7 (67,967 npm/wk · 70,976★ · v4.1.26) <https://cline.bot>
10. OpenCode: 54.0 (2,795,045 npm/wk · 222,610★ · v1.19.5) <https://opencode.ai>
11. OpenHands: 48.9 (92,083★ · v1.34.0) <https://openhands.dev>
12. Gemini CLI: 40.8 (288,968 npm/wk · 108,592★ · v0.66.0) <https://github.com/google-gemini/gemini-cli>
13. GitHub Copilot CLI: 36.4 (1,109,545 npm/wk · 11,305★ · v1.0.105) <https://github.com/github/copilot-cli>
14. Qwen Code: 32.2 (66,498 npm/wk · 28,771★ · sdk-typescript-v0.1.24) <https://github.com/QwenLM/qwen-code>
15. Aider: 30.5 (49,792★ · v0.86.0) <https://aider.chat>
16. Grok Build: 28.7 (70,176 npm/wk · 27,587★) <https://x.ai/build>
17. Crush: 26.8 (11,850 npm/wk · 28,845★ · v0.98.2) <https://github.com/charmbracelet/crush>
18. Goose: 25.5 (55,854★ · v1.56.0) <https://goose-docs.ai/>
19. Kimi CLI: 21.8 (11,423★ · 1.52.0) <https://github.com/MoonshotAI/kimi-cli>
20. Cursor CLI: 20.1 <https://cursor.com/docs/cli/overview>

## Most installed agent skills
Unit: installs, all time. Source: https://skills.sh/. Checked: 2026-10-28.

1. find-skills: 4.09M (vercel-labs/skills) <https://skills.sh/vercel-labs/skills/find-skills>
2. grill-me: 1.45M (mattpocock/skills) <https://skills.sh/mattpocock/skills/grill-me>
3. hyperframes-cli: 1.30M (heygen-com/hyperframes) <https://skills.sh/heygen-com/hyperframes/hyperframes-cli>
4. agent-browser: 1.28M (vercel-labs/agent-browser) <https://skills.sh/vercel-labs/agent-browser/agent-browser>
5. hyperframes: 1.26M (heygen-com/hyperframes) <https://skills.sh/heygen-com/hyperframes/hyperframes>
6. grill-with-docs: 1.24M (mattpocock/skills) <https://skills.sh/mattpocock/skills/grill-with-docs>
7. improve-codebase-architecture: 1.19M (mattpocock/skills) <https://skills.sh/mattpocock/skills/improve-codebase-architecture>
8. media-use: 1.17M (heygen-com/hyperframes) <https://skills.sh/heygen-com/hyperframes/media-use>
9. tdd: 1.16M (mattpocock/skills) <https://skills.sh/mattpocock/skills/tdd>
10. reddit-automation: 1.14M (flowkit-labs/skills) <https://skills.sh/flowkit-labs/skills/reddit-automation>
11. design-mobile-apps: 1.08M (designed-by-ai/skills) <https://skills.sh/designed-by-ai/skills/design-mobile-apps>
12. setup-matt-pocock-skills: 1.07M (mattpocock/skills) <https://skills.sh/mattpocock/skills/setup-matt-pocock-skills>
13. handoff: 1.06M (mattpocock/skills) <https://skills.sh/mattpocock/skills/handoff>
14. hyperframes-registry: 1.06M (heygen-com/hyperframes) <https://skills.sh/heygen-com/hyperframes/hyperframes-registry>
15. frontend-design: 1.04M (anthropics/skills) <https://skills.sh/anthropics/skills/frontend-design>

## Trending skills today
Unit: installs, last 24h. Source: https://skills.sh/trending. Checked: 2026-10-28.

Note: skills.sh trending includes skills-shell/superpowers entries with near-identical counts (~20.3K each) — possible install inflation; flag before ranking.

1. design-mobile-apps: 21.7K (designed-by-ai/skills) <https://skills.sh/designed-by-ai/skills/design-mobile-apps>
2. ui-taste: 21.4K (uizze.sh) <https://skills.sh/uizze.sh/ui-taste>
3. hyperframes-cli: 20.5K (heygen-com/hyperframes) <https://skills.sh/heygen-com/hyperframes/hyperframes-cli>
4. hyperframes: 20.2K (heygen-com/hyperframes) <https://skills.sh/heygen-com/hyperframes/hyperframes>
5. media-use: 19.9K (heygen-com/hyperframes) <https://skills.sh/heygen-com/hyperframes/media-use>
6. ai-image-generation: 19.8K (skills-shell/superpowers) <https://skills.sh/skills-shell/superpowers/ai-image-generation>
7. twitter-automation: 19.8K (skills-shell/superpowers) <https://skills.sh/skills-shell/superpowers/twitter-automation>
8. ai-video-generation: 19.8K (skills-shell/superpowers) <https://skills.sh/skills-shell/superpowers/ai-video-generation>
9. ai-avatar-video: 19.8K (skills-shell/superpowers) <https://skills.sh/skills-shell/superpowers/ai-avatar-video>
10. hyperframes-registry: 19.5K (heygen-com/hyperframes) <https://skills.sh/heygen-com/hyperframes/hyperframes-registry>

## Top OpenClaw skills (ClawHub)
Unit: downloads. Source: https://clawhub.ai/. Checked: 2026-10-28.

1. self-improving-agent: 489.2K (pskoett) <https://clawhub.ai/pskoett/skills/self-improving-agent>
2. skill-vetter: 280.1K (spclaudehome) <https://clawhub.ai/spclaudehome/skills/skill-vetter>
3. self-improving: 213.2K (ivangdavila) <https://clawhub.ai/ivangdavila/skills/self-improving>
4. github: 203.0K (steipete) <https://clawhub.ai/steipete/skills/github>
5. ontology: 200.9K (oswalpalash) <https://clawhub.ai/oswalpalash/skills/ontology>
6. gog: 198.2K (steipete) <https://clawhub.ai/steipete/skills/gog>
7. skillscan: 183.0K (tokauthai) <https://clawhub.ai/tokauthai/skills/skillscan>
8. proactive-agent: 178.5K (halthelobster) <https://clawhub.ai/halthelobster/skills/proactive-agent>
9. weather: 172.7K (steipete) <https://clawhub.ai/steipete/skills/weather>
10. multi-search-engine: 164.5K (gpyangyoujun) <https://clawhub.ai/gpyangyoujun/skills/multi-search-engine>

## Skill & agent template repos
Unit: GitHub stars. Source: https://github.com/. Checked: 2026-10-28.

1. openclaw/openclaw: 393.2K (OpenClaw agent (ClawHub client)) <https://github.com/openclaw/openclaw>
2. obra/superpowers: 306.9K (Superpowers skills/workflows) <https://github.com/obra/superpowers>
3. mattpocock/skills: 305.3K (Matt Pocock skills) <https://github.com/mattpocock/skills>
4. NousResearch/hermes-agent: 257.9K (Hermes Agent (bundled skills)) <https://github.com/NousResearch/hermes-agent>
5. anthropics/skills: 184.4K (Anthropic official skills repo) <https://github.com/anthropics/skills>
6. ComposioHQ/awesome-claude-skills: 78.2K (awesome list) <https://github.com/ComposioHQ/awesome-claude-skills>
7. heygen-com/hyperframes: 74.4K (HyperFrames skills) <https://github.com/heygen-com/hyperframes>
8. coreyhaines31/marketingskills: 57.0K (marketing skills) <https://github.com/coreyhaines31/marketingskills>
9. hesreallyhim/awesome-claude-code: 56.7K (awesome list (skills, hooks, commands)) <https://github.com/hesreallyhim/awesome-claude-code>
10. K-Dense-AI/scientific-agent-skills: 55.0K (scientific skills) <https://github.com/K-Dense-AI/scientific-agent-skills>
11. emilkowalski/skills: 50.9K (design/animation skills) <https://github.com/emilkowalski/skills>
12. anthropics/claude-plugins-official: 45.0K (official Claude Code plugins) <https://github.com/anthropics/claude-plugins-official>
13. vercel-labs/agent-browser: 44.7K (agent-browser skill/CLI) <https://github.com/vercel-labs/agent-browser>
14. wshobson/agents: 41.9K (subagents/skills collection) <https://github.com/wshobson/agents>
15. vercel-labs/skills: 37.0K (npx skills CLI) <https://github.com/vercel-labs/skills>

## MCP servers agents are wired to
Unit: public GitHub repos using it. Source: https://favz.co/census/tools.csv. Checked: 2026-10-28.

Note: Count of public GitHub repos whose agent config includes the server today (favz census).

1. @playwright/mcp: 5,130 <https://favz.co/>
2. url:mcp.supabase.com: 4,878 <https://favz.co/>
3. @upstash/context7-mcp: 4,703 <https://favz.co/>
4. shadcn: 2,214 <https://favz.co/>
5. url:api.githubcopilot.com: 2,007 <https://favz.co/>
6. chrome-devtools-mcp: 1,965 <https://favz.co/>
7. @modelcontextprotocol/server-github: 1,335 <https://favz.co/>
8. @modelcontextprotocol/server-filesystem: 1,271 <https://favz.co/>
9. @modelcontextprotocol/server-sequential-thinking: 1,203 <https://favz.co/>
10. next-devtools-mcp: 978 <https://favz.co/>
11. url:mcp.sentry.dev: 766 <https://favz.co/>
12. @angular/cli: 728 <https://favz.co/>
13. @modelcontextprotocol/server-memory: 712 <https://favz.co/>
14. url:mcp.figma.com: 648 <https://favz.co/>
15. @modelcontextprotocol/server-postgres: 622 <https://favz.co/>

## Free models right now
Unit: context window. Source: https://openrouter.ai/api/v1/models. Checked: 2026-10-28.

Note: OpenRouter ':free' variants: $0 in / $0 out, rate-limited. Free ends without notice.

1. Recursive: Flint: 262K ctx (recursive/flint-v1.0:free · added 2026-10-21 · tools) <https://openrouter.ai/recursive/flint-v1.0:free>
2. XingChen: Xing4.0 29B A4B: 262K ctx (xingchen-agi/xing4.0-29b-a4b:free · added 2026-10-16 · tools) <https://openrouter.ai/xingchen-agi/xing4.0-29b-a4b:free>
3. NVIDIA: Nemotron 3.5 Super VL: 1.0M ctx (nvidia/nemotron-3.5-super-vl:free · added 2026-10-15 · tools) <https://openrouter.ai/nvidia/nemotron-3.5-super-vl:free>
4. Dots Studio: Dots3-Note Preview: 512K ctx (dots-studio/dots-3-note-preview:free · added 2026-08-14 · tools) <https://openrouter.ai/dots-studio/dots-3-note-preview:free>
5. LiquidAI: LFM2.5-2.6B: 65K ctx (liquid/lfm-2.5-2.6b:free · added 2026-08-11 · tools) <https://openrouter.ai/liquid/lfm-2.5-2.6b:free>
6. NVIDIA: Nemotron 3.5 Lightning: 1.0M ctx (nvidia/nemotron-3.5-lightning:free · added 2026-08-11 · tools) <https://openrouter.ai/nvidia/nemotron-3.5-lightning:free>
7. Thinking Machines: Inkling Small: 1.0M ctx (thinkingmachines/inkling-small:free · added 2026-07-30 · tools) <https://openrouter.ai/thinkingmachines/inkling-small:free>
8. Poolside: Laguna S 2.1: 262K ctx (poolside/laguna-s-2.1:free · added 2026-07-21 · tools) <https://openrouter.ai/poolside/laguna-s-2.1:free>
9. Thinking Machines: Inkling: 1.0M ctx (thinkingmachines/inkling:free · added 2026-07-17 · tools) <https://openrouter.ai/thinkingmachines/inkling:free>
10. Poolside: Laguna XS 2.1: 262K ctx (poolside/laguna-xs-2.1:free · added 2026-07-02 · tools) <https://openrouter.ai/poolside/laguna-xs-2.1:free>
11. Cohere: North Mini Code: 256K ctx (cohere/north-mini-code:free · added 2026-06-17 · tools) <https://openrouter.ai/cohere/north-mini-code:free>
12. NVIDIA: Nemotron 3.5 Content Safety: 128K ctx (nvidia/nemotron-3.5-content-safety:free · added 2026-06-04) <https://openrouter.ai/nvidia/nemotron-3.5-content-safety:free>
13. NVIDIA: Nemotron 3 Ultra: 1.0M ctx (nvidia/nemotron-3-ultra-550b-a55b:free · added 2026-06-04 · tools) <https://openrouter.ai/nvidia/nemotron-3-ultra-550b-a55b:free>
14. NVIDIA: Nemotron 3 Nano Omni: 256K ctx (nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free · added 2026-04-28 · tools) <https://openrouter.ai/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free>
15. Google: Gemma 4 26B A4B : 262K ctx (google/gemma-4-26b-a4b-it:free · added 2026-04-03 · tools) <https://openrouter.ai/google/gemma-4-26b-a4b-it:free>
16. Google: Gemma 4 31B: 262K ctx (google/gemma-4-31b-it:free · added 2026-04-02 · tools) <https://openrouter.ai/google/gemma-4-31b-it:free>
17. Google: Lyria 3 Pro Preview: 1.0M ctx (google/lyria-3-pro-preview · added 2026-03-30) <https://openrouter.ai/google/lyria-3-pro-preview>
18. Google: Lyria 3 Clip Preview: 1.0M ctx (google/lyria-3-clip-preview · added 2026-03-30) <https://openrouter.ai/google/lyria-3-clip-preview>
19. Free Models Router: 200K ctx (openrouter/free · added 2026-02-01 · tools) <https://openrouter.ai/openrouter/free>

## New open-weight models (30 days)
Unit: Hugging Face likes. Source: https://huggingface.co/models?sort=trending. Checked: 2026-10-28.

1. DeepSeek-V4.1-Flash-1017: 3,480 (deepseek-ai · mit · 2026-10-17) <https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017>
2. flint: 1,123 (RecursiveSI · apache-2.0 · 2026-10-21) <https://huggingface.co/RecursiveSI/flint>
3. Spark-X2.6-4B: 833 (XHToken · apache-2.0 · 2026-10-19) <https://huggingface.co/XHToken/Spark-X2.6-4B>
4. Nanbeige4.3-33B: 699 (Nanbeige · apache-2.0 · 2026-10-24) <https://huggingface.co/Nanbeige/Nanbeige4.3-33B>
5. Underdog-Saluki-27B-1.0: 656 (ConwayResearch · apache-2.0 · 2026-10-08) <https://huggingface.co/ConwayResearch/Underdog-Saluki-27B-1.0>
6. LFM2.5-24B-A2B: 586 (LiquidAI · other · 2026-10-12) <https://huggingface.co/LiquidAI/LFM2.5-24B-A2B>
7. DeepSeek-V4.1-Flash-1017-UNCENSORED-FP8: 434 (dealignai · mit · 2026-10-18) <https://huggingface.co/dealignai/DeepSeek-V4.1-Flash-1017-UNCENSORED-FP8>
8. DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF: 419 (ISTA-DASLab · mit · 2026-10-23) <https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF>
9. Sage-A1.5: 352 (InternScience · apache-2.0 · 2026-10-15) <https://huggingface.co/InternScience/Sage-A1.5>
10. Xing4.0-29B-Intent: 323 (XingChen-AGI · apache-2.0 · 2026-10-15) <https://huggingface.co/XingChen-AGI/Xing4.0-29B-Intent>
11. Hy-Memory-30B-A3B: 317 (tencent · apache-2.0 · 2026-10-15) <https://huggingface.co/tencent/Hy-Memory-30B-A3B>
12. flint-GGUF: 303 (unsloth · apache-2.0 · 2026-10-20) <https://huggingface.co/unsloth/flint-GGUF>
13. LFM3-1.7B: 294 (LiquidAI · other · 2026-10-23) <https://huggingface.co/LiquidAI/LFM3-1.7B>
14. DeepSeek-V4.1-Flash-1017-abliterated-GGUF: 290 (orcarouter · mit · 2026-10-22) <https://huggingface.co/orcarouter/DeepSeek-V4.1-Flash-1017-abliterated-GGUF>
15. Beam-1.0-501B-A23B: 283 (reflection · apache-2.0 · 2026-10-07) <https://huggingface.co/reflection/Beam-1.0-501B-A23B>

## Best models by human votes (LMArena text)
Unit: Elo. Source: https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset. Checked: 2026-10-22.

1. Gemini 4 Argon (High): 1,526 (google · Proprietary) <https://lmarena.ai/leaderboard>
2. Claude Opus 5.5 (High): 1,516 (anthropic · Proprietary) <https://lmarena.ai/leaderboard>
3. Claude Fable 5.1 (Max): 1,513 (anthropic · Proprietary) <https://lmarena.ai/leaderboard>
4. Claude Opus 5 (Max): 1,508 (anthropic · Proprietary) <https://lmarena.ai/leaderboard>
5. Claude Opus 5 (High): 1,503 (anthropic · Proprietary) <https://lmarena.ai/leaderboard>
6. Claude Opus 4.6 (High): 1,501 (anthropic · Proprietary) <https://lmarena.ai/leaderboard>
7. Muse Wonder (Max): 1,501 (meta · Proprietary) <https://lmarena.ai/leaderboard>
8. Gemini 3.8 Flash (High): 1,498 (google · Proprietary) <https://lmarena.ai/leaderboard>
9. Claude Opus 4.6: 1,497 (anthropic · Proprietary) <https://lmarena.ai/leaderboard>
10. Muse Spark 1.3 (Max): 1,491 (meta · Proprietary) <https://lmarena.ai/leaderboard>
11. MiMo-V2.6-Pro: 1,491 (xiaomi · MIT) <https://lmarena.ai/leaderboard>
12. Claude Fable 5 (High): 1,491 (anthropic · Proprietary) <https://lmarena.ai/leaderboard>
13. Claude Opus 4.7 (High): 1,490 (anthropic · Proprietary) <https://lmarena.ai/leaderboard>
14. Gemini 3.7 Flash (High): 1,490 (google · Proprietary) <https://lmarena.ai/leaderboard>
15. Qwen3.8-Max: 1,484 (alibaba · Proprietary) <https://lmarena.ai/leaderboard>

## Cheapest & fastest way to run new open models
Unit: $ per 1M tokens, blended 3:1 in:out. Source: https://openrouter.ai/api/v1/models/{id}/endpoints · https://router.huggingface.co/v1/models. Checked: 2026-10-28.

Note: Cheapest = lowest blended price on OpenRouter (check quant + uptime). Speed = Hugging Face router live throughput; not every model has it.

- minimax/minimax-m3.1-flash: cheapest StreamLake $0.135/$0.54 per 1M (fp8), fastest novita 73 tok/s <https://openrouter.ai/minimax/minimax-m3.1-flash/providers>
- deepseek/deepseek-v4.1-flash-1017: cheapest Sail Research $0.101/$0.15 per 1M (fp4), fastest baseten 104 tok/s <https://openrouter.ai/deepseek/deepseek-v4.1-flash-1017/providers>
- reflection/beam-1.0-501b-a23b: cheapest Nebius $0.25/$1.0 per 1M (fp4), fastest baseten 299 tok/s <https://openrouter.ai/reflection/beam-1.0-501b-a23b/providers>
- prism-ml/ternary-bonsai-2-9b: cheapest Darkbloom $0.08/$0.13 per 1M (int4), no speed data <https://openrouter.ai/prism-ml/ternary-bonsai-2-9b/providers>
- moondream/moondream3-preview: cheapest PrimeIntellect $0.3/$2.5 per 1M (fp8), no speed data <https://openrouter.ai/moondream/moondream3-preview/providers>
- moondream/moondream3.1-9b-a2b: cheapest PrimeIntellect $0.3/$2.5 per 1M (fp8), no speed data <https://openrouter.ai/moondream/moondream3.1-9b-a2b/providers>
- meta/muse-spark-open: cheapest Makora $1.25/$4.25 per 1M (fp4), fastest baseten 113 tok/s <https://openrouter.ai/meta/muse-spark-open/providers>
- nvidia/nemotron-3.5-super-vl: cheapest Darkbloom $0.06/$0.35 per 1M (nvfp4), no speed data <https://openrouter.ai/nvidia/nemotron-3.5-super-vl/providers>

## Where agents find skills, tools & bots
Unit: . Source: see each row. Checked: 2026-10-28.

Note: Grok Bot (xAI harness, beta 2026-08-11) official template marketplace at x.ai/bot/marketplace plus community indexes; separate from Grok Build plugin marketplace and Grok Connectors.

1. skills.sh (Vercel): 1.52M (agent skills (SKILL.md) for Claude Code/Codex/Cursor/etc; installs leaderboard) <https://skills.sh/>
2. ClawHub: 489.2K (OpenClaw skills + plugins) <https://clawhub.ai/>
3. Hermes Skills Hub: 102.4K (aggregated skills (ClawHub, skills.sh, GitHub taps, LobeHub, browse.sh, NVIDIA)) <https://hermes-agent.nousresearch.com/docs/skills>
4. Smithery: 20.8K (MCP servers (hosted+local) with useCount) <https://smithery.ai/>
5. Official MCP Registry: &gt;30000 (MCP server metadata) <https://registry.modelcontextprotocol.io/>
6. Glama MCP: 111.6K (MCP servers) <https://glama.ai/mcp/servers>
7. Skillselion: 87.4K (skills 64,756; MCP 9,891; marketplaces 10,411; plugins 2,385) <https://skillselion.com/>
8. Monid: 2,968 (pay-per-call tool/API marketplace for agents ("OpenRouter for agent tools")) <https://monid.ai/>
9. browse.sh (Browserbase): 479 (site-specific browser-automation skills) <https://browse.sh/>
10. Favz: 189.1K (census of MCP/skills/plugins configured in public GitHub repos) <https://favz.co/>
11. Grok Bot Marketplace (official, xAI): 31 (Grok Bot templates; tabs: From Grok Bot Team, Engineering, Sales, Marketing, Design, Personal, Recruiting & People, Product, Operations) <https://x.ai/bot/marketplace>
12. Grok Bot Index: 489 (independent index of Grok Bot templates; "checked Oct 27, 2026") <https://grokbotindex.com/>
13. Grokdex: 89 (community ranked board of Grok Bot share links (votes/adds)) <https://grokdex.net/templates>
14. aiagents.org Grok Bot Hub: 278 <https://aiagents.org/grok-bots>
15. awesome-grok-bot: 41 <https://github.com/Anil-matcha/awesome-grok-bot>
16. Grok Build Plugin Marketplace: 309 (plugins (skills, commands, agents, hooks, MCP, LSP) for Grok Build CLI; launched 2026-06-11) <https://x.ai/news/grok-plugin-marketplace>
17. Grok Connectors:  <https://grok.com/connectors>
18. GPT Store / ChatGPT apps:  <https://chatgpt.com/gpts>
19. OpenRouter apps:  <https://openrouter.ai/apps>
20. Hugging Face Spaces:  <https://huggingface.co/spaces>
21. PulseMCP:  <https://www.pulsemcp.com/>
22. LobeHub skills: 505 <https://lobehub.com/skills>

## AI companies by revenue run-rate
Unit: USD per year. Source: per row (CNBC, Reuters, The Information, filings…). Checked: 2026-10-28.

Note: Latest public figure per company; dates and metric types differ (ARR, run-rate, quarterly ×4). Not strictly comparable.

1. Anthropic: $65.0B (frontier lab · 2026-07 · CNBC) <https://cnbc.com/2026/08/17/anthropic-says-annualized-revenue-climbed-to-65-billion-in-july.html>
2. OpenAI: $50.0B (frontier lab · 2026-09-30 · Quartz citing CNBC/FT) <https://qz.com/openai-annualized-revenue-50-billion-correction-ai-stocks-100826>
3. xAI / SpaceX 'Super Intelligence' segment: $30.4B (frontier lab (+ Cursor, cloud) · 2026-Q3 · Motley Fool on SpaceX Q3 results) <https://fool.com/investing/2026/10/26/spacex-now-trades-like-an-ai-stock-here-s-what-that-means-for-your-portfolio>
4. CoreWeave (public, CRWV): $10.3B (GPU cloud / infra · 2026-Q2 · The Register) <https://theregister.com/off-prem/2026/08/12/coreweave-revenue-doubles-as-debt-pile-reaches-356b/5286832>
5. Cursor (Anysphere): $4.0B (coding agent / IDE · 2026-06 · Dealroom.co) <https://app.dealroom.co/news/note/cursor-tops-4b-annualized-revenue-june-2026>
6. Mercor: $2.0B (AI training data / expert marketplace · 2026-06 · TechCrunch) <https://techcrunch.com/2026/07/08/these-ai-startups-are-growing-revenue-at-faster-and-faster-rates>
7. Surge AI: $1.4B (AI training data / RLHF · 2025-08 (reported May 2026) · aitrainer.work (cites Sacra/GetLatka)) <https://aitrainer.work/news/scale-ai-fallout-surge-ai-hyper-growth>
8. Together AI: $1.1B (inference / neocloud · 2026-Q2 · Value Add VC (cites company)) <https://valueaddvc.com/blog/how-does-together-ai-make-money-gpu-rentals-per-token-pricing-and-8-3b-valuation>
9. Fireworks AI: $1.0B (inference · 2026-07-15 · Fireworks (company blog)) <https://fireworks.ai/blog/series-d-announcement>
10. Modal Labs: $1.0B (serverless GPU / sandboxes infra · 2026-10-14 · Modal (company blog); SiliconANGLE) <https://modal.com/blog/1-billion-annualized-run-rate-series-d>
11. Cognition (Devin + Windsurf): $1.0B (coding agent · 2026-09 · Startup Fortune citing Bloomberg) <https://startupfortune.com/cognitions-devin-ai-coding-agent-doubles-revenue-to-1-billion-a-year>
12. Scale AI: $1.0B (AI training data · 2025 FY (stated May 2026) · Value Add VC citing Forbes (May 2026)) <https://valueaddvc.com/blog/scale-ai-valuation-what-the-data-labeling-company-is-worth-after-the-meta-deal>
13. Perplexity: $750M (AI search / answer engine · 2026-08 · Benzinga citing Reuters/The Information) <https://benzinga.com/markets/tech/26/08/61374579/nvidia-perplexity-30-billion-valuation-750-million-revenue-bezos>
14. Cerebras (public, CBRS): $720M (AI chips / inference cloud · 2026-Q2 · Cerebras IR) <https://investors.cerebras.ai/news-releases/news-release-details/cerebras-systems-fast-inference-cloud-business-nearly-quadruples>
15. Lovable: $600M (vibe-coding app builder · 2026-09 · Sacra (estimate)) <https://sacra.com/c/lovable>
16. Baseten: $600M (inference · 2026-03 · Sacra (estimate)) <https://sacra.com/c/baseten>
17. ElevenLabs: $600M (voice AI · 2026-09 · Value Add VC) <https://valueaddvc.com/blog/how-does-elevenlabs-make-money-api-enterprise-voice-agents-and-the-business-model-breakdown>
18. Vercel (v0): $600M (dev platform / app builder · 2026-10-01 · okara.ai citing The Information) <https://okara.ai/blog/vercel-revenue-growth>
19. Replit: $525M (coding agent / app builder · 2026-04 · Sacra (estimate)) <https://sacra.com/research/replit-passes-500m-year>
20. Midjourney: $500M (image/video generation · 2026 · Value Add VC (estimate)) <https://valueaddvc.com/company/midjourney>
21. Groq: $500M (inference chips / cloud · 2025 FY · Wikipedia) <https://en.wikipedia.org/wiki/Groq>
22. Mistral AI: $400M (frontier lab (EU) · 2026-02 · Presenc AI citing Financial Times) <https://presenc.ai/research/mistral-ai-revenue-and-arr-2026>
23. Harvey: $400M (legal AI · 2026-09 · SaaS Sentinel) <https://saassentinel.com/2026/09/10/harvey-raises-550m-at-15-5b-valuation-as-legal-ai-arr-tops-400m>
24. Glean: $300M (enterprise search / agents · 2026-05-28 · Business Wire (company)) <https://businesswire.com/news/home/20260528505530/en/Glean-Surpasses-%24300M-ARR-Unrivaled-Enterprise-Context-Fuels-AI-Adoption>
25. OpenEvidence: $300M (medical AI search · 2026-07 · Investing.com citing The Information) <https://investing.com/news/stock-market-news/openevidence-reaches-300m-revenue-run-rate-bank-of-america-says-93CH-4801016>
26. Suno: $300M (AI music · 2026-02 (reaffirmed 'far beyond' Oct 2026) · Digital Music News) <https://digitalmusicnews.com/2026/10/01/suno-revenue-october-2026>
27. Cohere: $240M (enterprise LLMs · 2025-12 · CNBC) <https://cnbc.com/2026/02/13/ai-startup-cohere-revenue-ipo.html>
28. Sierra: $200M (customer-service agents · 2026-05 · TradeOpus citing NYT) <https://tradeopus.com/stocks/sierra-reaches-200m-revenue-run-rate-as-ai-agents-gain-enterprise-budgets>
29. Runway: $200M (video generation / world models · 2026-09 · Clink citing Bloomberg) <https://clinkbill.com/arr-leaderboard/runway>
30. HeyGen: $200M (AI video avatars · 2026-06-25 · HeyGen (company)) <https://heygen.com/blog/heygen-surpasses-200m-arr>

## Open-weight labs: newest release
Unit: . Source: https://huggingface.co/. Checked: 2026-10-28.

1. Ant Group inclusionAI (Ling):  (Ling-3.1-flash-1019 · 2026-10-27 · check card) <https://huggingface.co/inclusionAI/Ling-3.1-flash-1019>
2. MiniMax:  (MiniMax-M3.1-Flash (428B MoE) · 2026-10-17 (HF), weights announced 2026-10-23 · Custom non-MIT license) <https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash>
3. DeepSeek:  (DeepSeek-V4.1-Flash-1017 · 2026-10-17 · MIT) <https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017>
4. IBM Granite:  (granite-switch-4.1-30b · 2026-10-14 · Apache-2.0 (typical)) <https://huggingface.co/ibm-granite/granite-switch-4.1-30b>
5. NVIDIA Nemotron:  (NVIDIA-Nemotron-3.5-Super-VL-120B-A12B; Nemotron 3.5 Bolt E4B (2026-10-23) · 2026-10-12 · NVIDIA Open Model License) <https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16>
6. Mistral AI:  (Mistral Large 4 (1T, 'Le Chonk') — API preview, weights announced but not on HF yet; latest LLM on HF: Mistral-Large-3-675B-Instruct-2512 · 2026-10-06 (ML4 announce); 2025-11-28 (Large 3 HF) · Apache-2.0 (Large 3)) <https://huggingface.co/mistralai/Mistral-Large-3-675B-Instruct-2512>
7. Xiaomi MiMo:  (MiMo-V2.6-Pro / Flash · 2026-09-27 · check card) <https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL>
8. StepFun:  (Step-5-Preview · 2026-09-20 · check card) <https://huggingface.co/stepfun-ai/Step-5-Preview>
9. Meta:  (Muse-Spark-Open-816B-A42B · 2026-09-15 · Custom Meta license) <https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B>
10. Zhipu / Z.ai (GLM):  (GLM-5.3 / GLM-5.3-Flash · 2026-08-25 · MIT) <https://huggingface.co/zai-org/GLM-5.3>
11. Alibaba Qwen:  (Qwen3.8-2.4T-A95B (LLM); Qwen3.8-Flash-Next (2026-08-24); Qwen-Image-2.1-Turbo (2026-10-09) · 2026-08-08 · Apache-2.0 for most; Qwen-Image-2.1-Turbo research license) <https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B>
12. Moonshot AI (Kimi):  (Kimi-K3 (2.8T MoE) · 2026-06-13 (HF repo) · Modified MIT) <https://huggingface.co/moonshotai/Kimi-K3>
13. Google Gemma:  (Gemma 4 (E2B/E4B/12B/26B-A4B/31B); QAT variants 2026-06 · 2026-04-02 · Apache-2.0) <https://huggingface.co/google/gemma-4-31B-it-qat-w4a16-ct>
14. OpenAI:  (gpt-oss-120b / gpt-oss-20b (no newer open LLM found) · 2025-08-04 · Apache-2.0) <https://huggingface.co/openai/gpt-oss-120b>

## On X this week
Unit: likes. Source: https://x.com/ (read-only search by our X researcher). Checked: 2026-10-28.

Note: Claims seen on X. Leads, not verified facts: a number only moves into the tables above after we check the official source.

1. @deepseek_ai: 43.3K (🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!  Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. W… · 2026-10-20) <https://x.com/deepseek_ai/status/2112476496283775135>
2. @deepseek_ai: 37.2K (🚀 DeepSeek-V4.1-Flash-1017 is here!  Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement … · 2026-10-17) <https://x.com/deepseek_ai/status/2111375640717033777>
3. @FireworksAI_HQ: 5,611 (DeepSeek V4.1 Flash 1017 runs at 1,521+ tok/s on Fireworks Ultrafast.  Fast enough to ship Tetris in 11 seconds, 7x faster than Claude Opus 5.5.  Join the Ultra… · 2026-10-21) <https://x.com/FireworksAI_HQ/status/2112973992097468795>
4. @ArtificialAnlys: 4,079 (Announcing Artificial Analysis Intelligence Index v5, featuring 2 new benchmarks and 3 upgrades. For the first time, the majority of the Index is built on priva… · 2026-10-28) <https://x.com/ArtificialAnlys/status/2115265133467635877>
5. @ImAI_Eruel: 3,182 (The benchmark scores for DeepSeek-V4.1-Flash-1017, which was just released, are turning out to be utterly ridiculous (with active parameters at 8B and 16B, it c… · 2026-10-17) <https://x.com/ImAI_Eruel/status/2111424490240675921>
6. @coldniko: 3,134 (Deepseek-V4.1-Flash now runs on 64GB of RAM with Strata.  Using GSQ-RCO quantizations by  @mhelcig · 2026-10-24) <https://x.com/coldniko/status/2113835370869583983>
7. @karanC_12: 2,707 (Opus 5.5 vs DeepSeek V4.1 Flash-1017   •Opus: 20 mins, $6 •DeepSeek: 1 hour, $0.23  Opus cooked, but DeepSeek got surprisingly close at 26x cheaper. · 2026-10-17) <https://x.com/karanC_12/status/2111428441019170937>
8. @HamelHusain: 2,299 (New blog post 🎉: How To Use Decision Models For Evals  In this post, we show you how to choose a decision model for your task, tune its threshold (don't assume … · 2026-10-20) <https://x.com/HamelHusain/status/2112588257104216294>
9. @SPAC89: 2,060 (GPT 6.1 Sol Ultra vs DeepSeek V4.1 Flash 1017 Max, I still can't believe a model that's 90% cheaper can produce results this good, how is this even possible? Bo… · 2026-10-17) <https://x.com/SPAC89/status/2111413922989224089>
10. @painn_x: 2,008 (WTFFF, this is absolutely insane 😳  deepseek v4.1 flash is free on opencode and tokenhabor right now  and i run it on deepseek's own agent harness for $0 no sub… · 2026-10-15) <https://x.com/painn_x/status/2110677445850325128>
11. @mhelcig: 1,764 (Run a 763B-parameter model locally. DeepSeek-V4.1-Flash-1017 on a 12GB GPU.  Frontier quality with GSQ-RCO quants, directly supported in the Strata inference en… · 2026-10-23) <https://x.com/mhelcig/status/2113769988813144118>
12. @deedydas: 1,681 (There is an absolute barrage of inference providers popping up that get to $1M/mo extremely quickly.  This is the rough economics of inference providers:  Even … · 2026-10-25) <https://x.com/deedydas/status/2114195317860266432>