Claude Opus 4.8 vs GPT-5.6 Terra
Wins 3 of 8 areas
Coding · Facts · Images and charts
Wins 4 of 8 areas
Agents · Reasoning · Math · Long documents
GPT-5.6 Terra wins more areas, narrowly.Claude Opus 4.8 is better at facts.
Both rank among the ten best models we track in coding.
Scores updated · 49 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- AgentsCarrying out multi-step tasks on its own03GPT-5.6 Terra3 of 5 tests · 2 ties
- MathCompetition and research-level math02GPT-5.6 Terra2 of 3 tests · 1 tie
- ReasoningHard problems that need careful thinking23GPT-5.6 Terra3 of 7 tests · 2 ties
- Long documentsFinding answers in very long texts01GPT-5.6 Terra1 of 1 test
- FactsGetting facts right instead of making them up30Claude Opus 4.83 of 3 tests
- Images and chartsUnderstanding pictures, charts and video20Claude Opus 4.82 of 2 tests
- CodingWriting and fixing software32Claude Opus 4.83 of 7 tests · 2 ties
- Following instructionsDoing exactly what it is asked11Even1 each
Long documents rests on a single test.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
GPT-5.6 Terra costs 53% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GPT-5.6 Terra pulls ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+13.7points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+12.2points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+11.8points ahead
Where Claude Opus 4.8 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+48.6points ahead
- Common-sense trick questionsSimpleBench+15.9points ahead
- Short factual questions, answered correctlySimpleQA Verified+9.8points ahead
Every test, side by side
All 49 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.8
- SWE-bench ProClaude Opus 4.8 by 5.869.263.4+5.8
- LiveBench · Agentic CodingGPT-5.6 Terra by 4.550.555+4.5
- LMArena · WebDevClaude Opus 4.8 by 37 rating points15561519+37 rating
- LiveBench · CodingClaude Opus 4.8 by 3.581.878.3+3.5
- Terminal-Bench 2.1GPT-5.6 Terra by 3.484.688+3.4
- Terminal-Bench Hardtie58.357.6tie
- SciCodetie54.455tie
AgentsGPT-5.6 Terra
- Terminal-Bench 4.0GPT-5.6 Terra by 13.721.735.4+13.7
- τ-Bench V3 · BankingGPT-5.6 Terra by 634.240.2+6
- BrowseCompGPT-5.6 Terra by 3.284.387.5+3.2
- AA ApexAgentstie39.438.9tie
- GDPValtie47.847.7tie
ReasoningGPT-5.6 Terra
- SimpleBenchClaude Opus 4.8 by 15.964.848.9+15.9
- ARC-AGI-2GPT-5.6 Terra by 11.872.183.9+11.8
- CritPtGPT-5.6 Terra by 9.120.930+9.1
- Humanity's Last ExamClaude Opus 4.8 by 5.848.742.9+5.8
- LiveBench · ReasoningGPT-5.6 Terra by 1.489.290.6+1.4
- ARC-AGI-3tie1.50.8tie
- GPQA Diamondtie9292.5tie
FactsClaude Opus 4.8
- AA-Omniscience · Non-hallucinationClaude Opus 4.8 by 48.660.712.1+48.6
- SimpleQA VerifiedClaude Opus 4.8 by 9.85343.2+9.8
- AA-Omniscience · AccuracyClaude Opus 4.8 by 248.846.8+2
Images and chartsClaude Opus 4.8
- CharXiv (RQ)Claude Opus 4.8 by 489.985.9+4
- LMArena · VisionClaude Opus 4.8 by 24 rating points12951271+24 rating
MathGPT-5.6 Terra
- FrontierMath Tier 4GPT-5.6 Terra by 12.256.168.3+12.2
- FrontierMath Tiers 1-3 (v2)GPT-5.6 Terra by 68086+6
- LiveBench · Mathematicstie94.394.9tie
Following instructionsEven
- IFBenchGPT-5.6 Terra by 962.271.2+9
- LiveBench · Instruction FollowingClaude Opus 4.8 by 7.47264.6+7.4
Other results19 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- GDP (Surge AI)Claude Opus 4.8 by 60.184.824.7+60.1
- AA-OmniscienceClaude Opus 4.8 by 28.728.80.1+28.7
- Agents' Last ExamGPT-5.6 Terra by 24.725.750.4+24.7
- livebench_data_analysisGPT-5.6 Terra by 13.36679.3+13.3
- DeepSWEGPT-5.6 Terra by 11.65869.6+11.6
- DeepSWE 1.1GPT-5.6 Terra by 115970+11
- τ²-Bench Telecom (AA run)Claude Opus 4.8 by 8.194.486.3+8.1
- ToolathlonClaude Opus 4.8 by 6.859.953.1+6.8
- GDPval-AA v2 EloClaude Opus 4.8 by 65 rating points15931528+65 rating
- OSWorld 2.0Claude Opus 4.8 by 5.555.750.2+5.5
- ARC-AGI-1GPT-5.6 Terra by 492.596.5+4
- livebench_languageGPT-5.6 Terra by 3.279.782.9+3.2
- Artificial Analysis Coding IndexGPT-5.6 Terra by 2.474.376.7+2.4
- HealthBenchClaude Opus 4.8 by 2.359.357+2.3
- Terminal-Bench 4.0Claude Opus 4.8 by 2.123.621.5+2.1
- HealthBench ProfessionalGPT-5.6 Terra by 1.955.857.7+1.9
- AA Agentic IndexGPT-5.6 Terra by 1.142.643.7+1.1
- AA Intelligencetie41.842.1tie
- Terminal-bench 3.0tie21.120.8tie
Questions people ask
Which is better, Claude Opus 4.8 or GPT-5.6 Terra?
GPT-5.6 Terra wins four of the eight areas we test: agents, reasoning, math and long documents. Claude Opus 4.8 wins coding, facts and images and charts. They are level on following instructions.
Which is better for coding?
Claude Opus 4.8. It wins 3 of the 7 coding tests both models report; GPT-5.6 Terra wins 2, and 2 are ties.
Which is cheaper?
Claude Opus 4.8 costs $5.00 per million input tokens and $25.00 per million output tokens; GPT-5.6 Terra costs $2.00 and $12.00. That makes GPT-5.6 Terra about 53% cheaper for the same work.
How do you compare the two?
We use the 49 benchmark tests both models have published scores on. The verdict counts the 30 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 19 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.