ModelWatch

Compare models

Up to 4 models side by side.

Provider
AnthropicOpenAIGoogle
Released
Nov 1, 2026seedAug 7, 2025seedJun 17, 2025seed
Input price
USD per 1M tokens
$4$1.25seed$1.25seed
Output price
USD per 1M tokens
$20$10seed$10seed
Workload cost
Chat reply (2K in, 500 out)
$0.02$0.0075seed$0.0075seed
Context window
1M400Kseed1.05Mseed
Max output
128K128Kseed66Kseed
Output speed
tokens per second
60seed70seed90seed
Time to first token
seconds
2sseed4sseed3sseed
GPQA Diamond
Graduate-level science questions that resist web search.
86%seed85%seed84%seed
SWE-bench Verified
Fixing real GitHub issues in Python repositories.
81%seed74%seed63%seed
MMLU-Pro
Broad knowledge and reasoning across 14 subjects.
87%seed86%seed86%seed
AIME
Competition math problems.
92%seed94%seed86%seed
Arena Elo
Human preference rating from head-to-head chats.
1505seed1460seed1455seed
Inputs
text, image, pdftext, imagetext, image, audio, video, pdf
Open weights
NoNoNo
Good at
codingagentswritingreasoninglong-context
reasoningcodingagents
long-contextvisionreasoningmultilingual
Notes
Anthropic's flagship everyday model. Strong at coding, agents and careful writing, at half the price of Fable.OpenAI's general-purpose reasoning model. Strong at math, coding and tool use.Google's long-context reasoning model with strong video and document understanding.

Best value in each row is highlighted. Share this page's URL to share the comparison.

Compare · ModelWatch