ChatGPT vs Gemini

Right now ChatGPT's flagship model is GPT-6 Astra and Gemini's is Gemini 4 Argon.

We pick flagships the same way as our subscription guide: the model with the highest overall score among those you can use on the plans. It changes when a new model comes out.

1 / 3
Swipe to compare

GPT-6 Astra is OpenAI's model for extended work involving computers, browsers, complex code and source material. OpenAI describes improvements over the preceding Sol in professional tasks, tool use and adherence to assigned scope, emphasizing continued execution after reviewing evidence. Connected runtimes provide software operation and file creation, combining the model's decisions about changes with external tools that carry them out. Development, research and document production can therefore form a workflow spanning applications, from gathering information through revising deliverables and checking results. The model detail page's specifications and evaluation analysis cover context and output limits, reasoning settings, computer-use and tool-call conditions, code repair, document analysis and multi-step automation.

Author
OpenAIOpenAI
Release Date
2026-09-04
Knowledge Cutoff
2026-04-30
License
Proprietary
I/O Format
Context Length
✦1.1M / 128K
API I/O (1M)
$10 / $50
How to Use
ChatGPT Plus (In Work and Codex) / ChatGPT Pro 100 or above / API Access
Overall Eval
Output Speed
✦37 tok/s
Arena Text
1475
Intelligence Index
✦56.9
Coding Index
77.1
Math Index
—
LiveBench
82.2
ForecastBench
—
Reasoning Eval
GPQA Diamond
96.3%
HLE
54.7%
MMLU-Pro
—
AIME 2025
—
MATH-500
—
LB Reasoning
92.7
LB Math
96.8
LB Data Analysis
83.0
Coding Eval
LiveCodeBench
—
LB Coding
80.4
LB Agentic
57.3
TAU2
—
TerminalBench
—
SciCode
56.5%
Language Eval
IFBench
—
AA-LCR
✦0.8
Hallucination (HHEM)
✦8.7%
Factual Consistency (HHEM)
✦91.3%
LB Language
89.4
LB Instruction Following
75.6

Gemini 4 Argon is Google's model for software development, long-document analysis and multi-step work involving text and visual material. Google says it expands the output available over earlier Gemini models, giving extended tasks more room for deliberation and finished deliverables. The company highlights code changes, specialist analysis and work across tools, alongside defenses against malicious instructions and actions outside the assigned scope. The combination connects extended deliberation with tool-driven changes and deliverable creation. Consult the model detail page's detailed specifications and evaluation analysis for context and output limits, reasoning settings, tool-call support, specialist document analysis and code repair.

Author
GoogleGoogle
Release Date
2026-09-30
Knowledge Cutoff
—
License
Proprietary
I/O Format
Context Length
1M / 1M
API I/O (1M)
✦$2 / $10
How to Use
Gemini AI Plus or above
Overall Eval
Output Speed
35 tok/s
Arena Text
✦1523
Intelligence Index
50.4
Coding Index
—
Math Index
—
LiveBench
—
ForecastBench
—
Reasoning Eval
GPQA Diamond
—
HLE
✦57.1%
MMLU-Pro
—
AIME 2025
—
MATH-500
—
LB Reasoning
—
LB Math
—
LB Data Analysis
—
Coding Eval
LiveCodeBench
—
LB Coding
—
LB Agentic
—
TAU2
—
TerminalBench
—
SciCode
✦61.8%
Language Eval
IFBench
—
AA-LCR
0.8
Hallucination (HHEM)
12.1%
Factual Consistency (HHEM)
87.9%
LB Language
—
LB Instruction Following
—

Top 10 models in the overall ranking

Updated 2026-11-08
See the full ranking
RankModelOverallAutomationDocs & officeCodingCost (per 1M tokens)Output speed
🥇Gemini 4 CarbonGoogle—92.1 tok/s
🥈Claude Opus 5.5Anthropic$16.0078.5 tok/s
🥉Claude Fable 5.1Anthropic$40.0065.0 tok/s
4Claude Fable 5Anthropic$40.0056.0 tok/s
5Gemini 4 ArgonGoogle$8.0035.0 tok/s
6GPT-6 AstraOpenAI$40.0036.5 tok/s
7GPT-6.1 SolOpenAI—$8.0041.5 tok/s
8Muse WonderMeta—$16.0092.0 tok/s
9Muse Spark 1.3Meta$3.5096.0 tok/s
10Claude Opus 5Anthropic$20.0065.0 tok/s

Latest AI model comparisons