GPT-5.2 Instant (source unspecified)

Name: GPT-5.2 Instant (source unspecified)
Author: OpenAI

OpenAI

This benchmark entry retains GPT-5.2 Instant observations whose source does not identify the evaluated version behind the mutable gpt-5.2-chat-latest route. The December 11, 2025 launch and February 10, 2026 update are separate documented versions; an alias, score publication date, or later card's latest-version comparison does not by itself identify either artifact.

Benchmark scores

Benchmark	Score
BullshitBench v2 Leaderboard	27.0
BuseyBench SVG	4.2
DTBench	81.3
FormationEval	97.4
HealthBench	55.4
HealthBench (length-adjusted)	50.6
HealthBench (updated implementation)	51.5
HealthBench Hard	26.8
HealthBench Hard (length-adjusted)	23.3
HealthBench Hard (updated implementation)	23.5
HealthBench Professional	38.3
HealthBench Professional (Length-Adjusted)	35.7
Hemingway-bench	994
LMCA	40.1
Ruby LLM Benchmarks — Program Fixer	72.1
Ruby LLM Program Fixer — Calendar	82.0
Ruby LLM Program Fixer — Parking Garage	51.8
Ruby LLM Program Fixer — School Library	76.7
Ruby LLM Program Fixer — Vending Machine	78.1
SnakeBench Average Apples	7.9
SnakeBench Best Apples	19.0
SnakeBench Rating	25.6
SnakeBench Total Apples	119
SnakeBench Win Rate	73.3
SpeciEval	90.7
SpeciEval — Belief in Animal Sentience	6.6
SpeciEval — Land Animal 4Ns	3.9
SpeciEval — Sea Animal 4Ns	4.3
SpeciEval — Speciesism	2.1
SpeciEval (Pooled English Runs)	90.7