Compare

Pick up to six models. You get a plain-language verdict, a spec sheet with the best value in each row highlighted, a chart, and every benchmark score side by side.

Claude Opus 5.5max×Gemini 4 Argonhigh×

Drag a model to change the order.

Verdict

• Claude Opus 5.5 (max) leads on quality: 72.7 vs 71.0.
• Claude Opus 5.5 (max) is strongest in coding, composite, knowledge, long context, maths, multimodal, reasoning.
• Gemini 4 Argon (high) is strongest in agents & tools, human preference.
• Gemini 4 Argon (high) is 2.0× cheaper than Claude Opus 5.5 (max) ($4.00 vs $8.00 per 1M blended).
• Gemini 4 Argon (high) starts answering in 47.4 s; Claude Opus 5.5 (max) takes 10.5 min.

Metric	Claude Opus 5.5 max Anthropic	Gemini 4 Argon high Google
Quality
BenchLeader Index50 = average model	72.7 — best of the models compared	71.0
Reasoning	75.7 — best of the models compared	74.9
Coding	82.2 — best of the models compared	72.1
Agents & tools	63.6	71.8 — best of the models compared
Maths	74.4 — best of the models compared	70.4
Knowledge	78.7 — best of the models compared	76.5
Human preference	–	72.5
Multimodal	71.0	–
Long context	67.5 — best of the models compared	64.9
Composite	94.2 — best of the models compared	88.5
Speed and latency
Output speed	95 tok/s — best of the models compared	83 tok/s
Time to first token	4.11 s — best of the models compared	4.23 s
Time to first answerafter any reasoning	628.2 s	47.4 s — best of the models compared
Context window	1M	1.0M — best of the models compared
Price
Input, per 1M tokens	$4.00	$2.00 — best of the models compared
Output, per 1M tokens	$20.00	$10.00 — best of the models compared
Blended (3 in : 1 out)	$8.00	$4.00 — best of the models compared
Cached input, per 1Mprompt-cache read price	$0.200	$0.100 — best of the models compared
Index per dollar	9.1	17.8 — best of the models compared
Cost and time per task
Chat reply: cost400 in / 300 out	$0.0076	$0.0038 — best of the models compared
Chat reply: cost with caching75% cache hits	$0.0065	$0.0032 — best of the models compared
Chat reply: time	10.5 min	51.0 s — best of the models compared
Summarize a 30-page report: cost12,000 in / 600 out	$0.060	$0.030 — best of the models compared
Summarize a 30-page report: cost with caching75% cache hits	$0.026	$0.013 — best of the models compared
Summarize a 30-page report: time	10.6 min	54.6 s — best of the models compared
Code edit: cost6,000 in / 1,500 out	$0.054	$0.027 — best of the models compared
Code edit: cost with caching75% cache hits	$0.037	$0.018 — best of the models compared
Code edit: time	10.7 min	1.1 min — best of the models compared
Agentic coding session: cost60,000 in / 4,000 out	$0.320	$0.160 — best of the models compared
Agentic coding session: cost with caching75% cache hits	$0.149	$0.074 — best of the models compared
Agentic coding session: time	11.2 min	1.6 min — best of the models compared
Structured extraction: cost2,000 in / 200 out	$0.012	$0.0060 — best of the models compared
Structured extraction: cost with caching75% cache hits	$0.0063	$0.0032 — best of the models compared
Structured extraction: time	10.5 min	49.8 s — best of the models compared

Chart

Quality againstPriceSpeedLatencyYour models are ringed and labelled; the rest is the frontier.

• Anthropic6
• Google3
• Meta3
• Alibaba2
• OpenAI2
• Reflection2
• Xiaomi2
• Ant Group1
• DeepSeek1
• IBM1
• MiniMax1
• NVIDIA1
• StepFun1
• Zhipu AI1

Every benchmark

Raw scores side by side, grouped by category. Best value in each row is purple. “Ref.” benchmarks are shown but not counted in the index.

Benchmark	Claude Opus 5.5 max	Gemini 4 Argon high
Reasoning
GPQA Diamond	90.6%	96.4% — best of the models compared
Arena Hard Prompts	–	1549
LiveBench 1.0 Reasoning (ref.)	92.2%	–
Humanity's Last Exam (AA) (ref.)	61.4% — best of the models compared	57.1%
ARC-AGI-1	97.5%	98.5% — best of the models compared
ARC-AGI-2	91.7% — best of the models compared	91.3%
ARC-AGI-3	–	97.3%
CritPt	31.7% — best of the models compared	27.1%
MysteryMechanism (ref.)	–	45.5%
Chess Puzzles (ref.)	67.0% — best of the models compared	59.0%
Mystery Game Puzzles (ref.)	71.0% — best of the models compared	64.0%
LMCA (ref.)	68.2% — best of the models compared	63.7%
DTBench (ref.)	98.9% — best of the models compared	97.3%
Coding
SciCode	66.9% — best of the models compared	61.8%
Arena Coding	–	1559
Arena WebDev	1807 — best of the models compared	1678
LiveBench 1.0 Coding (ref.)	89.3%	–
SciCode (AA) (ref.)	66.9% — best of the models compared	61.8%
IOI	–	100.0%
Code Migration (ref.)	–	68.2%
ProgramBench (ref.)	–	2.5%
Vibe Code Bench 1-100 (ref.)	–	21.2%
Vibe Code Bench v1.1 (ref.)	–	91.9%
MirrorCode (ref.)	77.4%	–
CursorBench (ref.)	57.8% — best of the models compared	52.9%
ALE-Bench (ref.)	–	2100.1
FrontierSWE (ref.)	62.3% — best of the models compared	55.0%
Terminal-Bench 4.0 (AA) (ref.)	59.6% — best of the models compared	57.1%
FrontierCode 1.1 (AA) (ref.)	52.2% — best of the models compared	43.5%
Agents & tools
Terminal-Bench	64.8%	–
APEX-Agents	73.5%	–
Arena Agent	20.9 — best of the models compared	12
LiveBench Agentic Coding (ref.)	71.7%	–
GDPval-AA v2.1 (ref.)	68.3% — best of the models compared	56.2%
ITBench SRE (AA) (ref.)	38.2%	–
Analyst Agent (AA) (ref.)	56.3% — best of the models compared	48.8%
CUA-bench (ref.)	14.0% — best of the models compared	4.8%
CyberBench (ref.)	–	77.9%
Finance Agent v2 (ref.)	–	65.4%
Harvey's Legal Agent Benchmark (ref.)	–	27.1%
Legal Research Bench (ref.)	–	54.8%
SREBench (ref.)	–	44.3%
Tax Agent Bench (ref.)	–	44.6%
Terminal-Bench 4.0 (Vals) (ref.)	–	57.6%
Terminal-Bench Science (ref.)	–	44.3%
Time Horizon Index: KSP (ref.)	–	52.5%
AutomationBench (ref.)	69.5%	77.5% — best of the models compared
GDP.pdf (AA) (ref.)	26.2% — best of the models compared	21.8%
Harvey LAB-AA v1.1 (ref.)	4.2% — best of the models compared	3.9%
AA-Briefcase v1.1 (ref.)	1803 — best of the models compared	1486
Terminal-Bench-Science 0.1 (AA) (ref.)	59.0% — best of the models compared	52.9%
CWE-Bench (AA)	58.3%	–
DeepSec-Bench (AA)	26.9%	–
CyberGym end-to-end (AA)	0.8%	–
Maths
FrontierMath Tiers 1–3	91.2% — best of the models compared	86.0%
FrontierMath Tier 4	95.0% — best of the models compared	78.0%
OTIS Mock AIME	100.0% — best of the models compared	100.0% — best of the models compared
ProofBench	100.0%	–
LiveBench Mathematics (ref.)	97.1%	–
Arena Maths	–	1529
Knowledge
SimpleQA Verified	72.2%	76.0% — best of the models compared
LiveBench 1.0 Data Analysis (ref.)	80.3%	–
AA-Omniscience	46.4 — best of the models compared	42.4
LegalBench	–	88.3%
BioMysteryBench (ref.)	–	76.3%
Excel Modeling Benchmark (ref.)	–	75.2%
MedCode (ref.)	–	58.8%
MedScribe (ref.)	–	87.4%
Public Benefits Bench (ref.)	–	69.8%
EBR-bench (ref.)	71.4% — best of the models compared	64.8%
GDP.pdf (ref.)	30.6% — best of the models compared	28.4%
AA-Omniscience: accuracy (ref.)	66.2% — best of the models compared	49.9%
AA-Omniscience: non-hallucination (ref.)	41.4%	84.9% — best of the models compared
Instruction following
LiveBench Language (ref.)	86.3%	–
LiveBench Instruction Following (ref.)	65.7%	–
Arena Instruction Following (ref.)	–	1525
Human preference
Arena Text	–	1524
Arena Creative Writing (ref.)	–	1516
Arena Multi-turn (ref.)	–	1552
Arena Longer Queries (ref.)	–	1543
Multimodal
MMMU-Pro	87.7%	–
SAGE (ref.)	–	53.6%
AudioMC (ref.)	–	58.4%
AudioMC (text output) (ref.)	–	58.4%
AA-3D (ref.)	1191 — best of the models compared	1154
Long context
AA-LCR	84.7% — best of the models compared	79.7%
MLCR (ref.)	27.8% — best of the models compared	13.9%
Composite
LiveBench 1.0 (ref.)	83.2%	–
AA Intelligence Index v4.3.2	57.6 — best of the models compared	52.6
Vals Index (ref.)	–	69.2
AA Cyber Index (ref.)	28.7	–
Safety & honesty
FORTRESS (ref.)	5.8%	–