Compare AI models
| Specifications | ||
Context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider. | 1M | 1.05M |
| API pricingUSD per 1M tokens · lower wins · base tier | ||
Input priceWhat you pay for everything you send the model — your question, plus any documents or earlier conversation you include with it. | $8.00 | $2.50 |
Output priceWhat you pay for the text the model writes back. It is normally the dearer half: producing an answer costs more than reading one. | $40.00 | $15.00 |
Cached input priceA reduced rate for text you send over and over. If every request starts with the same instructions or the same document, the provider keeps a copy ready and charges less to read it again. | $0.20 | $0.25 |
Cheapest inputLowest input rate across every provider on OpenRouter, the lab included. The cheapest endpoint may run a quantised build or a shorter context — see "Available from" on the model page. | — | $1.25OpenAI |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | — | $7.50OpenAI |
| $3,600Anthropic list price | $1,250OpenAI list price | |
| Benchmarks | ||
Humanity's Last Exam · with toolsMultidisciplinary reasoning — Humanity's Last Exam — extremely hard expert questions across many subjects. “With tools” means the AI is allowed to search the web or run code while answering. Higher is better. | 74.2% | 52.1% |
BullshitBench v2Nonsense detection — Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better. | — | 48% |
ProgramBenchProgram reconstruction — The AI receives a working program and its documentation, then builds a replacement from scratch without the original source code, internet access or decompilation. The score is the percentage of 200 programs that pass every behavioral test. We record each model's best published mini-SWE-agent result, including higher reasoning efforts where available. Partial test-pass rates and almost-solved programs do not count toward this score. Equal scores share a rank here; the official board also uses partial progress to break ties. Higher is better. | — | 0% |
SWE-Bench ProAgentic coding — Can the AI fix real bugs in real software? It's handed actual problems from open-source projects and has to write code that genuinely solves them. Higher is better. | — | 57.7% |
FrontierSWE V2Ultra-long-horizon coding — Engineering problems that would occupy a person for days: systems implementation, performance work, scientific computing and AI research, with up to twenty hours per task. Partial credit is awarded, because finishing one outright is still rare. Higher is better. | 69.9% | — |
FrontierCode 2.0Agentic coding — whether a model’s code changes to open-source repositories would be merged by their maintainers (version 2.0); separate from FrontierCode v1.1 scores | 49.1% | — |
Next.js EvalsNext.js coding — Vercel's open eval of how well AI coding agents build and migrate real Next.js apps — measured as the share of tasks the agent completes successfully. Higher is better. | — | 68% |
Terminal-Bench 2.0Agentic terminal coding — Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? (Version 2.0 of the test.) Higher is better. | — | 75.1% |
Terminal-Bench-Science 0.1Agentic scientific computing — The same command-line setup as Terminal-Bench, pointed at scientific work: the AI has to drive research tooling and computational workflows through to a result, rather than administer a machine. Version 0.1 is the first release of the task set, and scores run lower than on the general board. Higher is better. | 74.6% | — |
Expert-SWE (Internal)Software engineering — OpenAI's private set of expert-level software-engineering problems. Higher is better. | — | 68.5% |
MCP Atlas · original harnessMulti-step tool use — pass rate under the original tool-use harness, before Scale changed the judge and retry handling in April 2026 | — | 67.2% |
ToolathlonGeneral tool use — Tests how well the AI uses everyday real-world tools and apps to get things done. Higher is better. | — | 54.6% |
BrowseCompWeb browsing — Can the AI browse the web and track down hard-to-find answers? Higher is better. | — | 82.7% |
CyberGymCybersecurity — Tests the AI on cybersecurity challenges — finding and exploiting software weaknesses inside a safe sandbox. Higher is better. | — | 79% |
RoboDojo SimRobotics — control of a simulated two-arm robot from camera images, average success rate | 64.9% | — |
WANDR (relaxed score)Agentic search — agentic web search, relaxed scoring | 82.1% | — |
Humanity's Last Exam · no toolsMultidisciplinary reasoning — Humanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better. | — | 39.8% |
ARC-AGI-2Abstract reasoning — Puzzle-style tests of abstract reasoning and pattern-finding — the kind of thing people find easy but AIs often struggle with. Higher is better. | — | 73.95% |
FrontierMath · Tier 1–3Advanced math — Very hard, research-level math problems. Tiers 1–3 are the (still extremely difficult) lower tiers. Higher is better. | — | 47.6% |
FrontierMath · Tier 4Advanced math — Very hard, research-level math problems. Tier 4 is the hardest — close to what professional research mathematicians tackle. Higher is better. | — | 27.1% |
ARC-AGI-1Reasoning — accuracy on the first edition of the abstract grid reasoning benchmark | — | 93.7% |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | — | 92.8% |
OSWorld 2.1 · offlineAgentic computer use — Can the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Version 2.1, in its offline setting. Scores are not comparable with version 2.0. Higher is better. | 89.3% | — |
OSWorld-VerifiedAgentic computer use — Can the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Higher is better. | — | 75% |
AutomationBenchBusiness workflows — Tests whether the AI can run real multi-step business workflows — the kind of end-to-end office processes companies want to automate — from start to finish. Higher is better. | 51.4% | — |
FinanceAgent v1.1Finance — performance on financial research tasks in the v1.1 evaluation | — | 56% |
AA-Briefcase v1.1Knowledge work — Artificial Analysis agentic office-work eval (Elo, v1.1 rating fit) | 1785 | — |
AA-Briefcase v1.1 (rubric)Knowledge work — Artificial Analysis agentic office-work eval, v1.1 rubric score; separate from the Elo rating | 66.1% | — |
Remote Labor Index (RLI)Knowledge work — real paid remote-work projects, scored on the share completed to client standard | 41.3% | — |
GDPval (win/tie rate)Knowledge work — How often the AI's work matches or beats a human expert's on real knowledge-work tasks. Higher is better. | — | 83% |
Chartography · no toolsChart tasks — The same chart-centred test with no tools: the AI has to read each chart unaided. Scores run far lower than the with-tools version, so read the two as separate tests. Higher is better. | 73.4% | — |
MMMU-Pro · with toolsMultimodal reasoning — multimodal understanding and reasoning with access to tools | — | 82.1% |
MMMU-ProMultimodal reasoning — A tougher version of MMMU — college-level questions that mix images, diagrams, and text together. Higher is better. | — | 81.2% |
GraphWalks BFS (<128K)Long context — breadth-first traversal accuracy for inputs shorter than 128K tokens | — | 93% |
GraphWalks parents (<128K)Long context — parent-node retrieval accuracy for inputs shorter than 128K tokens | — | 89.8% |
MRCR v2 (8-needle) · 256K–512KLong context — mean match ratio for eight needles within the indicated context band | — | 57.5% |
MRCR v2 (8-needle) · 512K–1MLong context — mean match ratio for eight needles within the indicated context band | — | 36.6% |
| Overview | ||
| Company | Anthropic | OpenAI |
| Release date | Nov 9 2026 | Mar 5 2026 |
| Access | Closed | Closed |
| Model details | View model | View model |
Other comparisons
Claude Fable 5.2vsGemini 4 CarbonGPT-5.4vsGemini 4 CarbonClaude Fable 5.2vsMuse EmberGPT-5.4vsMuse EmberClaude Fable 5.2vsGrok 4.7GPT-5.4vsGrok 4.7Claude Fable 5.2vsDeepSeek-V4.1-Flash-1017GPT-5.4vsDeepSeek-V4.1-Flash-1017Claude Fable 5.2vsMistral Large 4GPT-5.4vsMistral Large 4Claude Fable 5.2vsKimi K3.1 ProGPT-5.4vsKimi K3.1 ProFrequently asked questions
Claude Fable 5.2 leads GPT-5.4 on 1 of the 1 benchmark they both report (Humanity's Last Exam). GPT-5.4 is cheaper on both input and output: $2.50 vs $8.00 per million input tokens, and $15.00 vs $40.00 per million output tokens. Figures are base-tier rates. GPT-5.4 shipped 249 days before Claude Fable 5.2, so benchmark comparisons should account for the intervening progress.
Context windows are 1M (Claude Fable 5.2) vs 1.05M (GPT-5.4).
On Humanity's Last Exam · with tools, Claude Fable 5.2 leads at 74.2% vs GPT-5.4 at 52.1%.