OpenAIReasoning modelCompare with another model

GPT-5.3-Codex

Reasoning effort

GPT-5.3-Codex is an OpenAI proprietary reasoning model, released 5 Feb 2026. Its best configuration (xhigh reasoning effort) ranks #57 of 775 on the BenchLeader Index at 63.8 ±5.0, in the upper half. The ± is the point: 155 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in agents & tools (80) and lowest in coding (61). At $4.81 per million tokens blended it is among the most expensive ranked models. Output speed of 120 tokens per second puts it faster than most, with a first answer in 58.0 s. It has been measured at 3 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 14 Oct 2026.

Blended price: $4.81/M; $1.75 in · $14.00 out
Output speed: 120 tok/s; measured by Artificial Analysis
First answer: 58 s; first token 1.98 s
Context: 400k
Full answer: 62 s; median, reasoning included
Cost per run: –
Released: 5 Feb 2026

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index64
Reasoning69
Coding61
Agents & tools80
Knowledge66
Instruction following74
Multimodal62
Long context67
Composite66

Versions

OpenAI has shipped 5 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.

Model	Released	Index	Rank
GPT-5.3-Codexxhighthis page	5 Feb 2026	63.8	#57
GPT-5.2-Codexxhigh	18 Dec 2025	59.4	#132
GPT-5.1-Codexhigh	12 Nov 2025	56.3	#208
GPT-5-Codexhigh	15 Sept 2025	57.5	#176
GPT 5.5 Codex	–	–	–

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
high	–	–	77 tok/s	1.98 s	$0.0049	Coding 56
xhighbest	63.8	#57	120 tok/s	58 s	$0.0049	Agents & tools 80 · Coding 61 · Composite 66 · Instruction following 74 · Knowledge 66 · Long context 67 · Multimodal 62 · Reasoning 69
not stated	–	–	77 tok/s	1.98 s	$0.0049	Agents & tools 58 · Coding 53

How its index has moved

19 Sept 2026 to 19 Oct 2026

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmark	xhigh	Source
GPQA Diamond (AA)not in index	91.5%#46	Artificial Analysis
Humanity's Last Exam (AA)not in index	42.5%#66	Artificial Analysis
CritPt	16.9%#81	Artificial Analysis

Coding

Benchmark	high	xhigh	not stated	Source
SWE-bench Verified (Epoch)	74.8%#15	–	–	Epoch AI Benchmarking Hub
WeirdML	–	77.9%#22	79.3%#19	WeirdML
Arena WebDeveffort not stated	–	–	1407#84	Arena (LMArena)
LiveCodeBench	–	87.3%#15	–	Vals AI
IOI	–	53.8%#21	–	Vals AI
SWE-bench (Vals)not in index	–	78.0%#33	–	Vals AI
Vibe Code Bench v1.1not in index	–	61.8%#51	–	Vals AI
ALE-Benchnot in index	–	1655.2#16	–	Epoch AI Benchmarking Hub

Agents & tools

Benchmark	high	xhigh	not stated	Source
Terminal-Bencheffort not stated	–	–	78.4%#6	Terminal-Bench
Terminal-Bench Hard	–	53.0%#17	–	Artificial Analysis
τ²-Bench Telecom (AA)not in index	–	86.0%#91	–	Artificial Analysis
HiL-Bencheffort not stated	–	–	4.3%#29	Scale AI SEAL
Terminal-Bench 2.0 (Vals)not in index	64.0%#6	–	–	Vals AI
METR Time Horizonsnot in indexeffort not stated	–	–	74.5%#6	METR
Vending-Bench 2not in indexeffort not stated	–	–	5940.1#22	Andon Labs

Knowledge

Benchmark	xhigh	Source
AA-Omniscience	10.9#92	Artificial Analysis
AA-Omniscience: accuracynot in index	52.9%#51	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	10.8%#400	Artificial Analysis

Instruction following

Benchmark	xhigh	Source	Trend
IFBench	75.4%#33	Artificial Analysis

Multimodal

Benchmark	xhigh	Source	Trend
MMMU-Pro	78.5%#72	Artificial Analysis

Long context

Benchmark	xhigh	Source	Trend
AA-LCR	83.3%#20	Artificial Analysis

Composite

Benchmark	xhigh	not stated	Source	Trend
Epoch Capabilities Indexnot in indexeffort not stated	–	157.1#24	Epoch AI Benchmarking Hub
AA Intelligence Index v4.3.2	32.5#110	–	Artificial Analysis

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
OpenAI	69 tok/s	1.98 s	$1.75	$14.00	400k	–
Azure	44 tok/s	5.15 s	$1.75	$14.00	400k	–

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.175 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0049	$0.0044	1.0 min
Summarize a 30-page report	12,000 / 600	$0.029	$0.015	1.1 min
Code edit	6,000 / 1,500	$0.032	$0.024	1.2 min
Agentic coding session	60,000 / 4,000	$0.161	$0.090	1.5 min
Structured extraction	2,000 / 200	$0.0063	$0.0039	59.7 s