Skip to main content
AI Choice Engine

Model comparison

GPT-5.3-Codex vs GPT-6 Sol

Two OpenAI tiers compared on the figures that decide which one a workload actually needs.

Catalog record checked September 26, 2026Individual provider fields may changeEvidence confidence: High — see receipts below

OpenAI

GPT-5.3-Codex

Frontier

vs

OpenAI

GPT-6 Sol

Balanced

AI model capability comparison
SpecificationGPT-5.3-CodexGPT-6 Sol
ProviderOpenAIOpenAI
TierFrontierBalanced
Context window400KWinner: 1.05M
Max output128K128K
Input / 1M tokensWinner: $1.75$2
Output / 1M tokens$14Winner: $10
WeightsClosedClosed
ParametersNot disclosedUnverifiedGPT-6 mid-price tier
Reasoning levelslow, medium, high, xhighnone, low, medium, high, xhigh, max
Modalitiestext, imagetext, image
API model idgpt-5.3-codexgpt-6-sol
ReleasedFebruary 5, 2026September 22, 2026
Artificial Analysis Intelligence Index [xhigh] (2026-09-26)32.5Not verifiedUnverified
Artificial Analysis Intelligence Index [max] (2026-09-26)Not verifiedUnverified47.5
DeepSWE 1.1 [max] (2026-09-22)Not verifiedUnverified68.8
OSWorld 2.0 (offline) [xhigh] (2026-09-22)Not verifiedUnverified60.5
Terminal-Bench 4.0 [max] (2026-09-26)Not verifiedUnverified43.9

Prices are USD per million tokens at standard rates, excluding batch and caching discounts. Bold indicates the better figure where one is objectively better. Values we could not confirm from the provider are shown as “Not verified” rather than estimated.

Benchmark receipts

Where each score comes from, and how far it can be compared across models.

  • 2026-09-26
    Artificial Analysis Terminal-Bench 4.0 (independent AA run, part of Intelligence Index v4.3.2)

    Agentic terminal work: long multi-step tasks executed in a sandboxed shell environment.

    Comparable with caveatNOT comparable with Terminal-Bench 2.x or 3.0 — 4.0 uses a new, non-overlapping task set. Within 4.0, scores from different harnesses (tbench.ai agent entries vs Artificial Analysis runs) are not interchangeable.

  • 2026-09-26
    Artificial Analysis

    Composite index blending reasoning, knowledge, and coding evals into one 0–100 score.

    Comparable with caveatOnly same-version scores are comparable. v4.3.2 was rebaselined (the top score fell from 66 on v4.1.1 to ~58), so v4.1.1 figures must not be compared with v4.3.2 figures — check measuredAt.

  • 2026-09-22
    OpenAI — Introducing GPT-6 Sol and Luna (vendor)

    Agentic software engineering: pass@1 on real repo tasks with reported cost per completed task.

    Directly comparable

Pricing tiers

GPT-5.3-Codex: Standard $1.75/$14 per MTok; cached input $0.175/MTok; Fast mode $3.50/$28. Still listed in the API as a coding-specialised model.

GPT-6 Sol: Replaces GPT-5.6 Sol at 50% of its $4/$20 promo price. Cached input $0.20/MTok; cache writes $2.50 (1.25x). Above 272K input the whole request bills $4/$15 ($0.40 cached). Batch/Flex $1/$5; Fast mode $4/$20. Knowledge cutoff Apr 20 2026. Launched in the API, ChatGPT Work and Codex the same day.

FrontierRecord checked September 26, 2026

GPT-5.3-Codex

GPT-5.3-Codex is OpenAI's coding-specialised API model at $1.75/$14 — the Codex agent itself now runs GPT-6 Sol, Luna and Astra.

Best for

  • Autonomous coding
  • Repo-scale refactors
  • Test generation

Watch out

Tuned for coding, not general chat; for new coding work compare GPT-6 Sol ($2/$10) first.

BalancedRecord checked September 26, 2026

GPT-6 Sol

GPT-6 Sol brings GPT-6 Astra's training advances to OpenAI's mid-priced tier — OpenAI claims roughly half GPT-5.6 Sol's factual-error rate at half its price, with a 1.05M-token context.

Best for

  • Default OpenAI production model
  • Agentic coding
  • Long-context knowledge work

Watch out

Brand-new (2026-09-22): the DeepSWE figure is vendor-reported and it is not yet on the independent boards; long prompts above 272K bill $4/$15 for the whole request.

When the cheaper one wins

GPT-6 Sol is cheaper on output at $10 per million tokens against $14 for GPT-5.3-Codex — about 1.4×. Use the cheaper tier for classification, extraction, summarisation, and any task where the expensive model’s extra score does not change the accepted output. The expensive one only pays if your hardest task actually fails on the cheap tier. These are standard-tier API rates, excluding batch and cache discounts.

Run the model picker

Evidence confidence: High

How strong and complete the evidence behind this comparison is — not a prediction of which model is better.

  • Pricing verified on both sides — Input and output rates are verified for both models.
  • 4/5 core specs verified on both sides — Not published for at least one side: parameter count.
  • 1 shared named benchmark with differing scores — Measured on: Artificial Analysis Intelligence Index.
  • Verified within the last 90 days — Newest catalog check was 2 days ago.
  • Both models carry source citations — Each side has at least two catalog sources on record.

Source receipts

Each catalog figure was checked against the provider or an independent second source on the date shown.

Related comparisons

Diving deeper on one model? GPT-5.3-Codex · GPT-6 Sol

Common questions

GPT-5.3-Codex vs GPT-6 Sol

Answered from the verified figures on this page rather than general guidance.

Is GPT-5.3-Codex or GPT-6 Sol cheaper for input?
GPT-5.3-Codex is cheaper at $1.75 per million input tokens, against $2 for GPT-6 Sol — roughly 1.1× the price. Output tokens usually dominate a real bill, so weigh the output rate more heavily than the input rate. These are base rates; GPT-5.3-Codex has tiered pricing: Standard $1.75/$14 per MTok; cached input $0.175/MTok; Fast mode $3.50/$28. Still listed in the API as a coding-specialised model. GPT-6 Sol has tiered pricing: Replaces GPT-5.6 Sol at 50% of its $4/$20 promo price. Cached input $0.20/MTok; cache writes $2.50 (1.25x). Above 272K input the whole request bills $4/$15 ($0.40 cached). Batch/Flex $1/$5; Fast mode $4/$20. Knowledge cutoff Apr 20 2026. Launched in the API, ChatGPT Work and Codex the same day.
Is GPT-5.3-Codex or GPT-6 Sol cheaper for output?
GPT-6 Sol is cheaper at $10 per million output tokens, against $14 for GPT-5.3-Codex — roughly 1.4× the price. Output tokens usually dominate a real bill, so weigh the output rate more heavily than the input rate. These are base rates; GPT-5.3-Codex has tiered pricing: Standard $1.75/$14 per MTok; cached input $0.175/MTok; Fast mode $3.50/$28. Still listed in the API as a coding-specialised model. GPT-6 Sol has tiered pricing: Replaces GPT-5.6 Sol at 50% of its $4/$20 promo price. Cached input $0.20/MTok; cache writes $2.50 (1.25x). Above 272K input the whole request bills $4/$15 ($0.40 cached). Batch/Flex $1/$5; Fast mode $4/$20. Knowledge cutoff Apr 20 2026. Launched in the API, ChatGPT Work and Codex the same day.
Which has the larger context window, GPT-5.3-Codex or GPT-6 Sol?
GPT-6 Sol accepts 1.05M tokens against 400K for GPT-5.3-Codex. This only matters if you routinely send very long documents or large codebases.
Do GPT-5.3-Codex and GPT-6 Sol support the same reasoning levels?
GPT-5.3-Codex exposes low, medium, high, xhigh, while GPT-6 Sol exposes none, low, medium, high, xhigh, max.
Should I use GPT-5.3-Codex or GPT-6 Sol?
GPT-5.3-Codex is the frontier tier and GPT-6 Sol the balanced tier. The useful question is whether your hardest task actually fails on the cheaper one — most production volume such as classification, extraction and summarisation does not.

Next step

Choosing between them

The verified figures that separate this pair, computed from the catalog rather than restated boilerplate.

  • Input price: GPT-5.3-Codex costs $1.75 per 1M tokens versus $2 for GPT-6 Sol — a 1.1x difference at the headline tier.
  • Context: GPT-6 Sol takes 1.05M against 400K for GPT-5.3-Codex — only decisive if your prompts approach the smaller window.
  • Positioning: GPT-5.3-Codex sits in the frontier tier, GPT-6 Sol in the balanced tier — most production volume (classification, extraction, summarisation) does not need the pricier tier.
Run the model pickerEstimate stack costAll model comparisons

Choosing a harness rather than a model — Cursor, Copilot, Claude Code, Windsurf, or Lovable? Compare agentic harnesses · Latest releases.