The Model Hub · updated September 19, 2026

Every model that matters.
None of the jargon.

Benchmark sites rank models for engineers. We answer the question everyone else is asking: what is it, and is it for me? Living pages, kept current as the labs ship.

8Model guides

5Head-to-heads

FreeNo signup, ever

This week ▸

Sep 17Antigravity Agent 09-2026 - Google refreshes its coding agentSep 15Gemini 3.8 Live - real-time speech-to-speech goes GASep 10DeepSeek-V4.1-Flash - 552B params, only 8B read your promptSep 03GPT-6 Astra - OpenAI’s new generation, $10 per million inSep 02Gemini 3.8 Flash ships - Google’s most capable Flash modelSep 01Claude Fable 5.1 - Anthropic’s new flagship, ahead of Opus 5Aug 26Qwen3.8-Flash-Next lands the day GLM-5.3-Flash doesAug 12Grok 4.6 - a 500K window for long-running agentsAug 12Qwen3.8-Max weights go public - 2.4T parameters, a first at this scaleAug 2026Muse Glimmer - Meta’s 30B open agent runs on one 24 GB cardPaceEight of these in 19 days - from the eight labs we trackSep 17Antigravity Agent 09-2026 - Google refreshes its coding agentSep 15Gemini 3.8 Live - real-time speech-to-speech goes GASep 10DeepSeek-V4.1-Flash - 552B params, only 8B read your promptSep 03GPT-6 Astra - OpenAI’s new generation, $10 per million inSep 02Gemini 3.8 Flash ships - Google’s most capable Flash modelSep 01Claude Fable 5.1 - Anthropic’s new flagship, ahead of Opus 5Aug 26Qwen3.8-Flash-Next lands the day GLM-5.3-Flash doesAug 12Grok 4.6 - a 500K window for long-running agentsAug 12Qwen3.8-Max weights go public - 2.4T parameters, a first at this scaleAug 2026Muse Glimmer - Meta’s 30B open agent runs on one 24 GB cardPaceEight of these in 19 days - from the eight labs we track

ChatGPT

The all-rounder everyone knows. Still the default assistant for most people - the smoothest app, voice and image tools built in, and the biggest ecosystem of integrations.

CTX 1M · Free - $200/mo→

Claude

The deep thinker for serious work. The model professionals quietly switch to.

CTX 1M · $20 - $200/mo→

Gemini

Lives inside everything Google. The most generous free tier of any frontier lab, the fastest release cadence in the industry - three Flash generations shipped between July and September 2026 alone - and it is already woven through Gmail, Docs and Android.

CTX 1M · Free - $200/mo→

Grok

Fast, irreverent, wired into X. The only assistant with a live feed of the world’s conversation.

CTX 500K · $8 - $40/mo→

DeepSeek

Frontier power at pocket prices. The model that made the whole industry cut prices, and it keeps doing it.

CTX 1M · Free→

Llama

The open one you can run yourself. The backbone of half the AI products you use without knowing it.

CTX Up to 1M (variant-dependent) · Free (open weights)→

Mistral

Europe’s efficient challenger. Small, sharp models that punch far above their size - and Europe’s only frontier lab, which matters if EU data residency is on your checklist.

CTX 256K · Free - $15/mo→

Qwen

The open powerhouse from the East. Quietly tops open-model leaderboards while the West looks elsewhere.

CTX 1M · Free (open weights)→

Head-to-head

Can’t pick between two? We already did the arguing.

Claude vs ChatGPT

ChatGPT is the better all-round app; Claude produces better work.

Gemini vs ChatGPT

Gemini wins on API price, release cadence and Google integration; ChatGPT wins on app polish and ecosystem.

Claude vs Gemini

Claude for output quality, Gemini for scale and price.

DeepSeek vs ChatGPT

DeepSeek delivers a startling share of ChatGPT’s capability for close to $0 - still the best value in AI.

Grok vs ChatGPT

Grok owns "right now" - real-time X grounding no rival matches.

Still not sure which one is yours?

Which AI is mine? →