GPT-5.5 Instant vs Claude Sonnet 4.6

Both mid-tier flagships at 90%+ — OpenAI vs Anthropic on real agentic tasks.

GPT-5.5 Instant 27/30 90.0% pass rate $0.0700/pass
vs
Claude Sonnet 4.6 28/30 93.3% pass rate $0.0514/pass Higher score Lower cost

Head-to-head: agentic-core-v1

Harness: agentic-core-v1 — 10 tasks × 3 runs = 30 total. Binary pass/fail per run. Harness version: openclaw@2026.4.22. Full methodology: what we measure and why.

Metric GPT-5.5 Instant Claude Sonnet 4.6
Runs passed / total 27 / 30 28 / 30
Pass rate 90.0% 93.3%
Campaign cost (30 runs) $1.89 $1.44
Cost per run $0.0630 $0.0480
Cost per passing run $0.0700 $0.0514
Provider OpenAI Anthropic
Campaign date 2026-05-10 2026-05-04

GPT-5.5 Instant

Strength

Clean execution on 9/10 task categories; confident short-path solving (avg 2 tool calls on task_09)

Weakness

task_09 wrong_answer all three runs — executes confidently on the wrong path without verification

Full campaign report →

Claude Sonnet 4.6

Strength

Consistent 9/10 task categories at 3/3; task_09 single-run pass via error-forced fallback

Weakness

$0.051/pass — 16× more expensive than Claude Haiku 4.5 for a 1-point score advantage

Full campaign report →
Cost ratio: Claude Sonnet 4.6 costs 1.4× less per passing run ($0.0514 vs $0.0700). At 10,000 tasks, that gap is $186 .

Verdict

Claude Sonnet 4.6 wins on both score (+1 run) and cost (1× cheaper per pass). It is the stronger choice across the board. The only reason to prefer GPT-5.5 Instant is ecosystem preference or tooling that requires OpenAI APIs specifically.

About agentic-core-v1

agentic-core-v1 is modelbattles' flagship benchmark harness. Ten tasks drawn from real engineering work: fix a failing test, refactor duplicated code, investigate a production log, add a null guard, trace through a codebase. Each task runs three times per model. A run passes if and only if the checker accepts the output — no partial credit, no manual grading.

The harness is deterministic: same task, same environment, same checker across all models. Scores are comparable. task_09 is the persistent difficulty point — it requires the model to recognise a structurally impossible calculation and refuse to produce a wrong answer instead of looping. Most models fail it at least once.

Read: What we actually measure and why · How to read an agentic-core-v1 score

More comparisons

ClawWorks Weekly

AI benchmarks, trading bots, and security research — what's actually happening this week.