AI models with code execution: they run the code, not guess it

Prices from official vendor price lists · October 2026
47 models that run code in a vendor sandbox while answering — so calculations are computed, not guessed. Output prices run from $0.3 to $180 per 1M tokens across the 47 models with a paid per-token price; 8 have a free tier and 3 ship open weights.
Code execution
Services found: 47
#AI modelCapabilitiesAPI $/1M (in / out)Context windowPrice / moOutput $/1MOverviewVisit site
41

GPT-5.2-Codex

OpenAI
Text & chatVisionCoding
$1.75 / $14400K tokensAPI only14View
42

GPT-5.1-Codex-Mini

OpenAI
Text & chatVisionCoding
$0.25 / $2400K tokensAPI only2View
43

GPT-5.1-Codex

OpenAI
Text & chatVisionCoding
$1.25 / $10400K tokensAPI only10View
44

GPT-5.5 Pro

OpenAI
Text & chatVisionReasoning
$30 / $1801.05M tokensAPI only180View
45

GPT-5.1-Codex-Max

OpenAI
Text & chatVisionCoding
$1.25 / $10400K tokensAPI only10View
46

GPT Audio Mini

OpenAI
Text & chatAudioStructured JSON output
$0.6 / $2.4128K tokensAPI only2.4View
47

GPT-5.4 Pro

OpenAI
Text & chatVisionReasoning
$30 / $1801.05M tokensAPI only180View

Browse by need

Reasoning models169 modelsCoding assistants17 modelsVision and images156 modelsLong context 1M+82 modelsFree tier available80 modelsOpen weights88 models

Forty-seven models can run code in a vendor-hosted sandbox while producing an answer - doing arithmetic properly, parsing an uploaded file, plotting a chart, testing a snippet before returning it. The sandbox is the vendor, so there is nothing to provision on your side.

Anthropic, OpenAI and Google all offer it, with output from $0.34 to $50 per million tokens. Sandbox time is billed separately from tokens, so check both meters before estimating cost. The capability matters most for data analysis: a model that calculates in a sandbox is reliably right, while a model calculating in its head is confidently approximate.

Computed, not guessedArithmetic and data work run in a real interpreter instead of being predicted.
Two metersTokens and sandbox time are billed separately; estimate both.
Files in, charts outThe sandbox can parse uploads and return generated images.

Models with this capability

The 47 models below all support this capability. Prices are input and output per 1M tokens from the vendor official price list; for dirhams multiply by 3.6725. Tap a card to open the model page.

Claude Fable 5.1
#1
Anthropic
in $10out $501M tokens
GPT-6 Astra
#2
OpenAI
in $10out $501.05M tokens
GPT-6.1 Sol
#3
OpenAI
in $2out $101.05M tokens
GPT-6 Sol
#4
OpenAI
in $2out $101.05M tokens
GPT-6 Luna
#5
OpenAI
in $0.1out $0.51.05M tokens
Claude Opus 5
#6
Anthropic
in $5out $251M tokens
Claude Fable 5
#7
Anthropic
in $10out $501M tokens
GPT-5.6 Sol
#8
OpenAI
in $4out $201.05M tokens

How to choose

  • Use it for anything numerical — model arithmetic remains unreliable without execution.
  • Check sandbox limits: memory, runtime and whether network access is available.
  • Watch session billing; short-lived containers are cheaper than long idle ones.
  • Validate outputs the same way you would validate a junior analyst.

Common mistakes

Trusting mental mathsA model without a sandbox will produce confident, wrong totals.
Leaving sessions openIdle containers bill by the minute on some platforms.

Who this suits

AnalyticsAd-hoc questions over uploaded spreadsheets.
FinanceCalculations that must be reproducible and checkable.
EngineeringTesting snippets before they reach a pull request.

Related filters

Coding17 models built for writing code
Tool calling166 models that call your functions
Structured JSON151 models with schema-guaranteed output
Reasoning169 models that think before answering
All 239 modelsThe full rating with filters and prices

Frequently asked questions

Where does the code run?

In a sandbox hosted by the vendor. Nothing to provision on your side.

Is sandbox time billed separately?

Yes. OpenAI charges per container session, Anthropic grants a monthly free allowance. Check both meters when estimating.

Why does it matter for analysis?

A model that calculates in a sandbox is reliably right; a model calculating in its head is confidently approximate.

Can the sandbox reach the internet?

Usually not by default — that is a deliberate safety boundary.

Which languages are supported?

Python dominates across vendors; some add shell access.

Frequently searched

By capabilityReasoningCodingVisionLong context 1M+Free tierOpen weightsLow latencyRealtime voiceImage generation
By vendorAnthropic (13)OpenAI (30)Google DeepMind (17)xAI (8)Meta (6)DeepSeek (9)Mistral AI (11)Alibaba (Qwen) (34)Z.ai (12)Moonshot AI (5)
Popular modelsClaude Fable 5.1GPT-6 AstraClaude Opus 5Claude Fable 5GPT-5.6 SolGLM 5.3Grok 4.6Kimi K3GPT-5.6 Terra
By deliveryBatch APIPrompt cachingPriority tierFrontier tierProprietary APITool callingStructured JSONWeb searchMultilingual

Ratings 0

Avg. score 0

Rate this page

Help us improve