Strong small GPT for coding subagents, quick tool use, and high-volume work
GPT-5.4 mini by openai is offered by 38 providers on this page. Public prices are shown for 36 of them. Its official list price is $0.75 per 1M input tokens and $4.50 per 1M output tokens. The lowest paid channel is Xpersona at $0.375 / $4.00 per 1M, about 2.5× 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 40.9 on the Intelligence Index, with 56.1 for coding and 31.5 for agentic tasks. Against its lowest blended price of $1.28 per 1M, that is roughly 32 index points per dollar, which is the value ratio the leaderboards rank on. Median output speed is 182 tokens per second, with 3.62s to the first token. Running one task of the Artificial Analysis suite costs about $0.4948, which reflects how many tokens its reasoning consumes rather than the unit price alone.
The context window is 400,000 tokens at the reference host, but hosts report different limits, from 128,000 to 400,000, 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. Providers report the capability flags inconsistently, so verify a specific feature against the host you plan to use. 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 |
|---|---|---|---|---|---|---|---|---|
| Kenari gpt-5-4-mini | Gateway | Free | — | — | 400,000 | 128,000 | ||
| Xpersona | Gateway | $0.375 | $4.00 | $0.037 | — | 272,000 ⚠ | 128,000 | |
| Poe | Gateway | $0.68 | $4.00 | $0.068 | — | 400,000 | 128,000 | |
| NanoGPT | Gateway | $0.75 | $4.50 | $0.075 | — | 400,000 | 128,000 | |
| Vivgrid | Gateway | $0.75 | $4.50 | $0.075 | — | 400,000 | 128,000 | |
| FreeModel | Gateway | $0.75 | $4.50 | $0.075 | $0.75 | 400,000 | 128,000 | |
| Impossibl | Gateway | $0.75 | $4.50 | $0.075 | — | 400,000 | 128,000 | |
| AIHubMix | Gateway | $0.75 | $4.50 | $0.075 | — | 400,000 | 128,000 | |
| OrcaRouter | Gateway | $0.75 | $4.50 | $0.075 | — | 400,000 | 128,000 | |
| Azure Cognitive Services | First-party | $0.75 | $4.50 | $0.075 | — | 400,000 | 128,000 | |
| OpenAIOfficial | First-party | $0.75 | $4.50 | $0.075 | — | 400,000 | 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.
1 more channels offer $0 (Kenari); free tiers usually have rate limits and no SLA, excluded from ranking.
| Name | Conditions | Score | Metric | Source |
|---|---|---|---|---|
| SWE-Bench Pro | variant: reasoning effort xhigh | 54.4 | resolve rate | Source ↗ |
| Terminal-Bench | variant: reasoning effort xhigh · v2.0 | 60 | accuracy | Source ↗ |
| MCP Atlas | variant: reasoning effort xhigh | 57.7 | score | Source ↗ |
| Toolathlon | variant: reasoning effort xhigh | 42.9 | score | Source ↗ |
| τ²-Bench Telecom | variant: reasoning effort xhigh | 93.4 | accuracy | Source ↗ |
| GPQA Diamond | variant: reasoning effort xhigh | 88 | accuracy | Source ↗ |
| Humanity's Last Exam | variant: with tools | 41.5 | accuracy | Source ↗ |
| Humanity's Last Exam | variant: without tools | 28.2 | accuracy | Source ↗ |
| OSWorld-Verified | variant: reasoning effort xhigh | 72.1 | success rate | Source ↗ |
| MMMU Pro | variant: with Python | 78 | accuracy | Source ↗ |
| MMMU Pro | variant: without tools | 76.6 | accuracy | Source ↗ |
| OmniDocBench | variant: reasoning effort none · v1.5 | 0.1263 | overall edit distance | Source ↗ |
| OpenAI MRCR | variant: 8-needle, 64K-128K · vv2 | 47.7 | accuracy | Source ↗ |
| OpenAI MRCR | variant: 8-needle, 128K-256K · vv2 | 33.6 | accuracy | Source ↗ |
| Graphwalks | variant: BFS, 0-128K | 76.3 | accuracy | Source ↗ |
| Graphwalks | variant: parents, 0-128K | 71.5 | accuracy | Source ↗ |
The same benchmark scores very differently across harness / dataset, so the qualifying conditions must be shown together.
1 / 38 providers expose no reasoning control (reasoning_options: []).
| Mode | Provider | Input | Output | Cache read | Cache write |
|---|---|---|---|---|---|
| fast | AIHubMix | $1.50 | $9.00 | $0.15 | — |
| fast | OrcaRouter | $1.50 | $9.00 | $0.15 | — |
| fast | Databricks | $1.50 | $9.00 | $0.15 | — |
| fast | Merge Gateway | $1.50 | $9.00 | $0.15 | — |
| fast | OpenAI | $1.50 | $9.00 | $0.15 | — |