B.O.B. Leaderboard

Anthropicreleased 28 May 2026

Claude Opus 4.8

#18 of 9538 of 46 benchmarksConsistent

68.7

B.O.B.

Category profile

  • Real use (hard to game)2×69
  • Reasoning & knowledge57
  • Math72
  • Coding66
  • Agents & real work66
  • Independent indexes81

Reality check

Lab tests

65

Real use (9 tests)

69

Lab tests and real-use tests agree. It does 4 points better on tests that are hard to train for than on the usual lab tests.

What counts as real use

Price per 1M tokens (OpenRouter)

$5.00 in · $25.00 out

Cost per task on the AA index

—

Output speed (Artificial Analysis)

—

§Receipts

Every benchmark result

“Beats” = the share of tracked models this one beat on that test. The raw score is its best reported setting.

BenchmarkRawBeatsSource
Real use (hard to game) · Real use 69
Kagi LLM Benchmarkas claude-opus-4-8-thinking88.8%
93%of 22
Kagi
LMArena Agent Arenaas Claude Opus 4.8 (High)0.075
80%of 50
LMArena (real agent sessions)
LMArena Textas claude-opus-4-8-high1482
75%of 80
LMArena (blind human votes)
LiveBench 2.0 (real-work suites)as claude-opus-4.876.4%
74%of 39
LiveBench / Abacus.AI
Vals Indexas anthropic/claude-opus-4-855.1%
68%of 42
Vals AI (private test sets)
EQ-Bench Creative Writingas claude-opus-4-81840
62%of 46
Sam Paech (open source)
SWE-rebenchas Claude Opus 4.8-xhigh56.5%
61%of 34
Nebius
LMArena WebDevas claude-opus-4-8-high1556
59%of 82
LMArena (blind human votes)
SimpleBenchas claude-opus-4-8_unknown64.8%
53%of 30
Epoch AI Benchmarking Hub
Reasoning & knowledge · Reasoning 57
Chess Puzzlesas claude-opus-4-8_max52.0%
81%of 73
Epoch AI Benchmarking Hub
CritPt (physics research)as claude-opus-4-8_max20.9%
75%of 89
Epoch AI Benchmarking Hub
Mystery Game Puzzlesas claude-opus-4-8_max36.0%
70%of 57
Epoch AI Benchmarking Hub
SimpleQA Verifiedas claude-opus-4-8_max53.0%
66%of 57
Epoch AI Benchmarking Hub
GPQA Diamondas claude-opus-4-8_max91.0%
66%of 70
Epoch AI Benchmarking Hub
ARC-AGI-2as claude-opus-4-8_high72.1%
55%of 44
Epoch AI Benchmarking Hub
EnigmaEvalas claude-opus-4-8_xhigh23.5%
44%of 10
Epoch AI Benchmarking Hub
ForecastBenchas claude-opus-4-8_24K59.8%
41%of 18
Epoch AI Benchmarking Hub
Blueprint-Bench 2as claude-opus-4-8_unknown14.5%
17%of 31
Epoch AI Benchmarking Hub
Math · Math 72
FrontierMath Tiers 1-3as claude-opus-4-8_max80.0%
75%of 60
Epoch AI Benchmarking Hub
ProofBenchas claude-opus-4-8_max69.0%
75%of 68
Epoch AI Benchmarking Hub
OTIS Mock AIMEas claude-opus-4-8_max98.3%
72%of 73
Epoch AI Benchmarking Hub
FrontierMath Tier 4as claude-opus-4-8_max56.1%
66%of 52
Epoch AI Benchmarking Hub
Coding · Coding 66
WeirdMLas claude-opus-4-8_xhigh82.9%
82%of 52
Epoch AI Benchmarking Hub
ALE-Benchas claude-opus-4-8_high1563.8%
77%of 71
Epoch AI Benchmarking Hub
FrontierCodeas claude-opus-4-8_unknown46.5%
71%of 45
Epoch AI Benchmarking Hub
GSO (code optimisation)as claude-opus-4-8_unknown47.1%
60%of 11
Epoch AI Benchmarking Hub
SciCodeas claude-opus-4-8_max53.5%
59%of 86
Epoch AI Benchmarking Hub
DeepSWEas claude-opus-4-8_max59.0%
48%of 26
Epoch AI Benchmarking Hub
Agents & real work · Agents 66
Game Boy Advance evalas claude-opus-4-8_unknown70.9%
91%of 23
Epoch AI Benchmarking Hub
LMCAas claude-opus-4-8_max57.5%
88%of 67
Epoch AI Benchmarking Hub
OSWorld 2.0as claude-opus-4-8_max20.6%
75%of 9
Epoch AI Benchmarking Hub
DTBenchas claude-opus-4-8_max94.9%
73%of 67
Epoch AI Benchmarking Hub
GDP.pdfas claude-opus-4-8_max24.0%
63%of 37
Epoch AI Benchmarking Hub
DeepResearch Benchas claude-opus-4-8_high50.2%
57%of 8
Epoch AI Benchmarking Hub
Vending-Bench 2as claude-opus-4-8_unknown5787.4%
52%of 47
Epoch AI Benchmarking Hub
Remote Labor Indexas claude-opus-4-8_unknown8.3%
50%of 9
Epoch AI Benchmarking Hub
APEX-Agentsas claude-opus-4-8_max48.9%
43%of 48
Epoch AI Benchmarking Hub
Independent indexes · Indexes 81
Epoch Capabilities Indexas Claude Opus 4.8158.3
81%of 81
Epoch AI