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. | $2.00 | $2.00 |
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. | $10.00 | $10.00 |
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. | $0.10 | $0.20 |
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. | $1.00Google AI Studio | $1.00OpenAI |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | $5.00Google AI Studio | $5.00OpenAI |
| $900Google list price | $900OpenAI list price | |
| Benchmarks | ||
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. | 66% | 59% |
DeepSWE 1.1Agentic coding — Artificial Analysis' independent test of deep, agentic software-engineering work — the AI has to plan and carry out substantial coding tasks end to end. (Version 1.1 of the test.) Higher is better. | 77.9% | 68.8% |
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. | 100% | 100% |
Terminal-Bench 4.0Agentic terminal coding — Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Version 4.0 recalibrated how much time, CPU and memory each task gets, removed eight tasks and fixed nineteen, so fewer runs fail for reasons that have nothing to do with the model. Scores are not comparable with earlier versions. Higher is better. | 57.4% | 49.39% |
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. | 91.25% | 89.58% |
OSWorld 2.0Agentic computer use — Can the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Version 2.0 is a harder, refreshed task set. Higher is better. | 69.2% | 60.5% |
Agent's Last Exam · pass@1Agentic computer use — A hard set of desktop and operating-system tasks an AI agent has to finish by looking at the screen and working the machine itself. The score is the share it passes outright — partial credit does not count. Higher is better. | 39.5% | 56.4% |
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. | 51.3% | 33.2% |
threejsevalCommunity preference (Three.js) — Every model gets the same prompt — "the Eiffel Tower", "a glass fishbowl", "a robot arm picking toys into a box" — and builds a 3D scene in Three.js. Real people then see two scenes side by side, names hidden, and vote for the one they prefer. The votes become a chess-style Elo rating on threejseval.com, averaged across all the prompts. It measures whether the scene looks and moves right to a human eye, not whether the code passes a test. Higher is better. | 1714 | 1622 |
CheatBenchCheating rate — Measures how often AI agents try to cheat on difficult assignments, such as reading hidden answers, copying work or manipulating grading. The overall score gives equal weight to ten categories; the sycophancy category measures how far an agent shifts its beliefs toward a user's stated views. Scores describe each model in its tested agent setup, and task success is measured separately. Lower is better. | — | 71.9% |
Gray Swan IPI · k = 15Prompt injection robustness — Attackers hide malicious instructions inside content the AI reads — a web page, an email, a document — and try to hijack what it does. This variant gives the attacker 15 tries and counts an attack as successful if any of them works. Lower is better. | 0.7% | — |
CursorBench 4.0Agentic coding — Cursor's own test of coding agents on ambiguous, multi-file tasks taken from real Cursor sessions — editing, refactoring, investigating a codebase, understanding what the user meant, managing jobs and following a design. Cursor runs each model at several reasoning efforts; each release here carries the score of its best listed effort. Scores aren't comparable with earlier CursorBench versions. Higher is better. | 52.9% | — |
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. | 55% | — |
Supabase Evals · with skillsSupabase coding — Supabase's own open benchmark: a coding agent is dropped into a real Supabase project and asked to do real work — set up a schema, fix a broken security policy, debug an Edge Function — and every run is checked against a live Supabase stack. This is the headline number, where the agent has Supabase's own skills loaded, as most people building on Supabase would. The score is the share of scenarios it got right. Higher is better. | — | 78.3% |
Supabase Evals · no skillsSupabase coding — The same Supabase scenarios, but with none of Supabase's skills loaded — so it measures what the model already knows about building on Supabase, rather than how well it follows Supabase's supplied instructions. Higher is better. | — | 78.3% |
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. | 57.6% | — |
RoboDojo SimRobotics — control of a simulated two-arm robot from camera images, average success rate | 34.2% | — |
RiemannBenchAdvanced mathematics — mathematical problem-solving evaluation published by Surge | 76% | — |
LAB-Bench 2Biology — Everyday tasks from a working biology lab — reading protocols, interpreting figures and sequence data, and answering the practical questions a researcher hits at the bench. Higher is better. | 88.8% | — |
CWE-bench v1Security patching — Collinear AI’s programmatically graded security-patch pass@1 rate; separate from exploit-generation benchmarks | 68% | — |
Finance Agent v2Agentic financial analysis — Tests the AI on real financial-analysis work, like digging through reports and making sound decisions. Higher is better. | 65.4% | — |
Harvey's Legal Agent BenchmarkAgentic legal work — Harvey's test of whether an AI agent can complete real legal work — drafting and reviewing documents, working with spreadsheets and presentations, and navigating files the way a lawyer's assistant would. Higher is better. | 19.6% | — |
HealthBench · full setHealthcare — physician rubric score on the full set of the original HealthBench evaluation | — | 47.1% |
HealthBench · full set, length-adjustedHealthcare — physician rubric score on the full set of the original HealthBench evaluation, adjusted for answer length | — | 53.2% |
HealthBench · hardHealthcare — physician rubric score on the hard subset of the original HealthBench evaluation | — | 22.1% |
HealthBench · hard, length-adjustedHealthcare — physician rubric score on the hard subset of the original HealthBench evaluation, adjusted for answer length | — | 30.1% |
HealthBench · consensusHealthcare — physician rubric score on the consensus subset of the original HealthBench evaluation | — | 95.8% |
HealthBench · consensus, length-adjustedHealthcare — physician rubric score on the consensus subset of the original HealthBench evaluation, adjusted for answer length | — | 96.2% |
HealthBench Professional · OpenAI grading, rawHealth — unadjusted physician-rubric score using the official OpenAI grading procedure | — | 59.5% |
HealthBench Professional · OpenAI grading, length-adjustedHealth — official OpenAI grading with the response-length penalty applied | — | 60.8% |
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. | 71.6% | — |
LVBenchVideo understanding — Can the AI follow a very long video — up to an hour — and answer questions that need details from far apart in it? Higher is better. | 91.7% | — |
GraphWalks BFS F1 (up to 128K)Long context — breadth-first search F1 on 650 problems with up to 128K tokens of context | 99.7% | — |
GraphWalks BFS (256K–1M)Long context — breadth-first search F1 on the 200 GraphWalks problems with 256K to 1M tokens of context | 84.2% | — |
| Overview | ||
| Company | OpenAI | |
| Release date | Sep 30 2026 | Sep 22 2026 |
| Access | Closed | Closed |
| Model details | View model | View model |
Other comparisons
Gemini 4 ArgonvsClaude Fable 5.2GPT-6 SolvsClaude Fable 5.2Gemini 4 ArgonvsMuse EmberGPT-6 SolvsMuse EmberGemini 4 ArgonvsGrok 4.7GPT-6 SolvsGrok 4.7Gemini 4 ArgonvsDeepSeek-V4.1-Flash-1017GPT-6 SolvsDeepSeek-V4.1-Flash-1017Gemini 4 ArgonvsMistral Large 4GPT-6 SolvsMistral Large 4Gemini 4 ArgonvsKimi K3.1 ProGPT-6 SolvsKimi K3.1 ProFrequently asked questions
Gemini 4 Argon leads GPT-6 Sol on 7 of the 9 benchmarks they both report. Both charge $2.00 per million input tokens. Both charge $10.00 per million output tokens. Figures are base-tier rates. GPT-6 Sol shipped 8 days before Gemini 4 Argon, so benchmark comparisons should account for the intervening progress.
Context windows are 1M (Gemini 4 Argon) vs 1.05M (GPT-6 Sol).
On BullshitBench v2, Gemini 4 Argon leads at 66% vs GPT-6 Sol at 59%. On DeepSWE 1.1, Gemini 4 Argon leads at 77.9% vs GPT-6 Sol at 68.8%. On Next.js Evals, both models score 100%. On Terminal-Bench 4.0, Gemini 4 Argon leads at 57.4% vs GPT-6 Sol at 49.39%. On ARC-AGI-2, Gemini 4 Argon leads at 91.25% vs GPT-6 Sol at 89.58%. On OSWorld 2.0, Gemini 4 Argon leads at 69.2% vs GPT-6 Sol at 60.5%. On Agent's Last Exam · pass@1, GPT-6 Sol leads at 56.4% vs Gemini 4 Argon at 39.5%. On AutomationBench, Gemini 4 Argon leads at 51.3% vs GPT-6 Sol at 33.2%. On threejseval, Gemini 4 Argon leads at 1714 vs GPT-6 Sol at 1622.