Agent-ready GPT for coding and computer-use workflows at a lower cost
GPT-5.4 by openai is offered by 45 providers on this page. Public prices are shown for 41 of them. Its official list price is $2.50 per 1M input tokens and $15.00 per 1M output tokens. The lowest paid channel is Xpersona at $0.75 / $6.00 per 1M, about 5.0× below the list price. That channel is a third-party gateway, so confirm its availability and rate limits before depending on it. It also has 1 free ($0) channel; free tiers usually carry rate limits, and subscription-covered access bills $0 per token only after the subscription fee.
Artificial Analysis rates it 39 on the Intelligence Index, with 71.1 for coding. Against its lowest blended price of $2.06 per 1M, that is roughly 19 index points per dollar, which is the value ratio the leaderboards rank on. Median output speed is 91 tokens per second, with 154.53s to the first token.
The context window is 1,050,000 tokens at the reference host, but hosts report different limits, from 400,000 to 1,050,000, so the usable window depends on the provider you pick. It supports reasoning, tool use, and structured output. Accepted input modalities are Text, Image, and PDF. Its training knowledge cuts off at 2025-08-31.
Quality is independently evaluated by Artificial Analysis. Speed/latency are model-level medians.
| Provider | Tier | Input | Output | Cache read | Cache write | Context | Output limit | Status |
|---|---|---|---|---|---|---|---|---|
| UnoRouter gpt-5.4:free | Gateway | Free | — | — | 1,050,000 | 128,000 | ||
| Xpersona | Gateway | $0.75 | $6.00 | $0.075 | — | 1,050,000 | 128,000 | |
| UnoRouter | Gateway | $1.80 | $10.80 | — | — | 1,050,000 | 128,000 | |
| Ofox | Gateway | $2.00 | $12.00 | $0.20 | — | 1,050,000 | 128,000 | |
| Poe | Gateway | $2.20 | $14.00 | $0.22 | — | 1,050,000 | 128,000 | |
| Perplexity Agent | Gateway | $2.50 | $15.00 | $0.25 | — | 1,050,000 | 128,000 | |
| NanoGPT | Gateway | $2.50 | $15.00 | $0.25 | — | 1,050,000 | 128,000 | |
| NEAR AI Cloud | Gateway | $2.50 tiered >272K: $5.00 | $15.00 | $0.25 | — | 1,050,000 | 128,000 | |
| Databricks databricks-gpt-5-4 | Cloud | $2.50 tiered >272K: $5.00 | $15.00 | $0.25 | — | 1,050,000 | 128,000 | |
| Kilo Gateway | Gateway | $2.50 | $15.00 | $0.25 | — | 1,050,000 | 128,000 | |
| OpenAIOfficial | First-party | $2.50 tiered >272K: $5.00 | $15.00 | $0.25 | — | 1,050,000 | 128,000 | |
Sorted by blended price (input×0.75 + output×0.25) asc. The official channel always shows regardless of rank. Whether a gateway's low price is actually usable can't be verified.
Interleaved thinking (reasoning between tool calls) is declared by 3 of 45 providers.
3 / 45 providers expose no reasoning control (reasoning_options: []).
| Mode | Provider | Input | Output | Cache read | Cache write |
|---|---|---|---|---|---|
| fast | Databricks | $5.00 | $30.00 | $0.50 | — |
| fast | OrcaRouter | $5.00 | $30.00 | $0.50 | — |
| fast | AIHubMix | $5.00 | $30.00 | $0.50 | — |
| fast | Merge Gateway | $5.00 | $30.00 | $0.50 | — |
| fast | OpenAI | $5.00 | $30.00 | $0.50 | — |
| fast | Snowflake Cortex | $5.00 | $30.00 | $0.50 | — |
| fast | AnyAPI | $5.00 | $30.00 | $0.50 | — |
1 more channels offer $0 (UnoRouter); free tiers usually have rate limits and no SLA, excluded from ranking.
| Name | Conditions | Score | Metric | Source |
|---|---|---|---|---|
| SWE-Bench Pro | dataset: public | 59.1 | resolve rate | Source ↗ |
| SWE-Atlas Codebase QnA | harness: Codex | 40.8 | score | Source ↗ |
| SWE-Atlas Codebase QnA | harness: Mini-SWE-Agent | 36.3 | score | Source ↗ |
| SWE-Atlas Refactoring | harness: Codex | 44.29 | score | Source ↗ |
| SWE-Atlas Test Writing | harness: Codex CLI | 44.36 | score | Source ↗ |
| SWE-Atlas Test Writing | harness: Mini-SWE-Agent | 40 | score | Source ↗ |
| Artificial Analysis Coding Agent Index | harness: Codex · variant: medium | 53.6 | average pass@1 | Source ↗ |
| SWE-Atlas Codebase QnA | harness: Codex · variant: medium | 72.4 | pass@1 | Source ↗ |
| SWE-Bench Pro | harness: Codex · variant: medium · dataset: hard-aa | 18.4 | pass@1 | Source ↗ |
| Terminal-Bench | harness: Codex · variant: medium · v2.1 | 69.8 | pass@1 | Source ↗ |
| Artificial Analysis Coding Agent Index | harness: Cursor CLI · variant: medium | 52.2 | average pass@1 | Source ↗ |
| SWE-Atlas Codebase QnA | harness: Cursor CLI · variant: medium | 72.9 | pass@1 | Source ↗ |
| SWE-Bench Pro | harness: Cursor CLI · variant: medium · dataset: hard-aa | 18.9 | pass@1 | Source ↗ |
| Terminal-Bench | harness: Cursor CLI · variant: medium · v2.1 | 64.7 | pass@1 | Source ↗ |
| Terminal-Bench | v2.0 | 75.1 | success rate | Source ↗ |
| GPQA Diamond | — | 92.8 | accuracy | Source ↗ |
| Humanity's Last Exam | variant: no tools | 39.8 | accuracy | Source ↗ |
| Humanity's Last Exam | variant: with tools | 52.1 | accuracy | Source ↗ |
| OSWorld-Verified | — | 75 | success rate | Source ↗ |
| BrowseComp | — | 82.7 | accuracy | Source ↗ |
| GDPval | — | 83 | wins or ties | Source ↗ |
| ARC-AGI-2 | variant: Verified | 73.3 | accuracy | Source ↗ |
| FrontierMath | dataset: Tier 1-3 | 47.6 | accuracy | Source ↗ |
| FrontierMath | dataset: Tier 4 | 27.1 | accuracy | Source ↗ |
| MMMU Pro | variant: no tools | 81.2 | accuracy | Source ↗ |
The same benchmark scores very differently across harness / dataset, so the qualifying conditions must be shown together.
Individual evaluations run by Artificial Analysis, on the same reasoning tier as the intelligence score above. Each benchmark has its own task set and harness, so rows are not comparable with one another. The Intelligence Index above draws on a different, newer set of evaluations.