Specs, pricing, context windows and plain-English verdicts for the frontier models people are actually choosing between.
A compact index for model selection: maker, fit, price posture and launch window.
The "Opus-class" workhorse, released 9 July — $2/$6 per 1M tokens, 500K context, ~4x fewer output tokens per task than Opus 4.8.
Workhorse · Jul 2026Full-duplex voice models that listen while they speak — now powering ChatGPT Voice, delegating hard questions to GPT-5.5. API "soon".
Voice · Jul 2026Anthropic's most capable released model — pulled under US export controls, back globally 1 July. $10/$50, 1M context.
Frontier · Jun 2026Mid-tier with 1M context by default and a real agentic step up. Intro pricing $2/$10 through 31 Aug, then $3/$15.
Mid-tier · Jun 2026OpenAI's new flagship family, GA 9 July — 1.05M context, $5/$30 (Sol), split benchmark record vs Claude Fable 5.
Flagship · Jul 2026Current Opus flagship — same $5/$25 rate card as 4.7, better benchmarks, fast mode now 3x cheaper at $10/$50.
Flagship · May 2026Beats Gemini 3.1 Pro on coding at 25% lower cost — $1.50/$9, 1M context, 76.2% Terminal-Bench 2.1.
Production · May 2026Alibaba's proprietary top model — ~1T-param MoE, 1M context, built for hours-long agent runs at $2.50/$7.50.
Flagship · May 2026Pro & Turbo variants behind China's #1 AI chatbot (155M WAU). Agent-focused, ¥6/¥30 per 1M tokens.
Agents · Jun 2026Reasoning, vision and coding unified in one 119B MoE with just 6B active params per token.
Efficient · Mar 2026Sonnet 4-level coding at one-third the cost and 2x+ the speed. $1/$5, 200K context, the subagent workhorse.
Budget · Oct 2025First full retrain since GPT-4.5. 1M context, $5/$30 pricing, 82.7% on Terminal-Bench 2.0. Superseded by GPT-5.6 at the same price.
Prev gen · Apr 20262M-token context — the largest available. Native video, audio and text in one pass, no transcription step.
Multimodal · Apr 2026Largest open-weights model of 2026 — 1.6T params, MIT-licensed, 80.6% SWE-bench, from $0.14/1M tokens.
Open weights · Apr 2026Open Apache 2.0 family — four sizes from phone to workstation, 89.2% on AIME, runs on one GPU.
Open source · Apr 2026Open-weight 1T-param coding model — ties GPT-5.5 on SWE-bench Pro at ~80% lower cost.
Coding · Apr 2026AI code editor rebuilt around agent orchestration — the new Agents Window. Shipped 2 Apr 2026.
Dev tool · Apr 202687.6% on SWE-bench Verified. 1M context, $5/$25 pricing. Superseded by Opus 4.8 at the same price.
Prev gen · Apr 2026Reportedly 6T parameters, trained on Colossus 2. Q1 and Q2 2026 targets both missed — Grok 4.5 shipped instead; still no date.
Unreleased · TBDCalculators built directly on the model index — same data, zero guesswork.
Enter monthly input/output token volume and compare the bill across Claude, GPT, Gemini, DeepSeek, Qwen and Kimi — from published rate cards.
Free tool · Jul 2026Can your GPU run it? Parameters × quantization + KV cache by context length → estimated VRAM, checked against 4090, 3090, 16GB cards and Mac unified memory.
Free tool · Jul 2026Paste text, get real GPT token counts (o200k/cl100k in your browser), labelled Claude/Gemini estimates, and the cost of that text per model.
Free tool · Jul 202638 X-vs-Y pages across Claude, GPT, Gemini, Grok, DeepSeek, Qwen and Kimi — real spec tables, price tables and published benchmarks.
Compare hubThe pairings that decide real architecture choices: quality margin, context, cost and deployment fit.
Half the price and a genuine benchmark split — Sol leads Terminal-Bench, Fable 5 leads SWE-bench Pro.
Updated Jul 2026Same $5/$30 price, higher published scores — the straightforward upgrade path.
Updated Jul 2026A near-exact SWE-bench Pro tie (64.6% vs 64.7%) next to a 2.5-5x price gap.
Updated Jul 2026The "Opus-class" claim, tested — a 2-2 benchmark split at a 2.5-4x price gap.
Updated Jul 2026Mid-tier challenger vs shipping flagship — a 2.5-3x price gap and matching 1M contexts.
Updated Jul 2026Anthropic's two top models at a clean 2x price gap — when does the Fable premium pay off?
Updated Jul 2026Same price, better benchmarks, 3x cheaper fast mode — the upgrade math, explained.
Updated Jul 2026The new Flash beats 3.1 Pro on coding at 25% lower cost. Where 3.1 Ultra still wins.
Updated Jul 2026Which flagship wins on coding, context window, multimodal ability and price?
Updated May 2026Open MIT weights vs closed flagship — an 8-10x price gap and who each suits.
Updated May 2026Code quality vs agentic reliability — Opus leads SWE-bench Pro, GPT-5.5 owns Codex.
Updated May 2026Two open-weight coding models head to head — context, price and licensing.
Updated May 2026Code quality vs multimodal scale — 1M vs 2M context, and a 2x price gap.
Updated May 2026Every pairing with complete published data — pricing, context window and specs side by side.
New AI models ship almost every week, each with its own benchmarks, pricing tiers and marketing language. AI Model Hub cuts through it: every page gives you the specs, the pricing, the benchmark numbers and a plain-English read on who the model is actually for.