GPT-5.4
OpenAI · released Mar 5, 2026
- Type
- Proprietary
- Input
- $2.50
- Output
- $15.00
- Cached
- $0.25
List price · per 1M tokens · OpenAI at 1.1M context · source ↗
Our take
Written Aug 3, 2026GPT-5.4 is OpenAI's flagship reasoning model that can handle over one million tokens in a single request, accepting text, images and files. It is available only through approved providers' APIs, with a wide spread of speed and cost tiers from budget batch work to premium real-time inference.
Choose this for long-document analysis at full-book scale, or for coding where it holds the top measured score in its profile. Use the budget tier for batch work, or premium tiers when speed matters. Skip it if you need to self-host, run air-gapped, or fine-tune; if creative writing quality is central; or if you want the cheapest frontier option.
The case for it
- Extremely long request limit for a frontier-class model: 1,050,000 tokens supports full books or large codebases in one pass.
- Measured coding performance leads its own benchmark profile, with an Arena Coding Elo 28–87 points above its own text scores.
- Wide speed/cost trade-off within the same model: a 5.3× throughput difference between the slowest and fastest tiers.
- Multiple provider options create genuine pricing pressure, with one host offering 24% faster throughput than OpenAI's own tier at the same per-token cost.
The case against it
- Creative writing is a relative soft spot, scoring 22–42 points below its own text scores and 78–87 points below its coding peak.
- Math performance shows evaluation variance and trails its own coding and hard-prompt scores by 38–61 points.
- Proprietary weights prevent local deployment, air-gapped use, or fine-tuning; all access requires API-only hosting through approved providers.
How good is it?
IntelligencePuzzles, maths, exam questions
Arena Text (overall)24th of 143 · 1465.4
Also on this board: 1477 via High (Aug 2, 2026). Read the pair, not the higher one.
CodingWriting and fixing code on its own
Arena Coding21st of 143 · 1514
Also on this board: 1521.5 via High (Aug 2, 2026). Read the pair, not the higher one.
AgenticPlanning, calling tools, staying on task
Arena Agent (IPS)11th of 36 · 0.051via High
WritingWe do not rate this
Two boards come close and neither tests writing: Arena Creative Writing asks people which of two replies they prefer, and LiveBench Language tests whether a model understood a passage. So we show where GPT-5.4 placed and give it no mark out of five.
These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done, which is why they get no rating.
Every published score for this model16 scoresEvery figure we hold, from 16 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Or rent it from someone else
Why this differs from the header. The strip above quotes OpenAI's own list price. This is the cheapest live offer at the widest standard context we hold, whoever is serving it — a reseller undercutting a lab is ordinary commerce, not an error.
- per 1M tokens
- $2.50 in / $15.00 out
- Context served
- 1.1M
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenAIflex tier | $1.25 / $7.50 | 1.1M | 27 tok/s | No | Yesunknown period | Unknown |
| Microsoft Azure AI | $2.50 / $15.00 | 1.1M | 56 tok/s | No | No | Confirmed |
| OpenRouter | $2.50 / $15.00 | 1.1M | not measured | Unknown | Unknown | Unknown |
| OpenAI | $2.50 / $15.00 | 1.1M128K out | 45 tok/s | No | Yesunknown period | Unknown |
| Microsoft Azure AIeu | $2.75 / $16.50 | 1.1M | 55 tok/s | No | No | Confirmed |
| Amazon Bedrockus-east-1 | $2.75 / $16.50 | 1.1M | 142 tok/s | No | No | Unknown |
| OpenAIpriority tier | $5.00 / $30.00 | 1.1M | 61 tok/s | No | Yesunknown period | Unknown |
Microsoft Azure AI
Amazon BedrockAcross the 7 listings we hold: 6 say they do not train on prompts, 0 say they do and 1 do not say. 2 appear in the zero-retention registry we check; the rest are unknown to us rather than confirmed either way.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
- ✓
- Supported
- ✗
- Not supported
- Not published
- host gave no parameter list
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| OpenAIflex | ✓ | ✓ | ✓ |
| Microsoft Azure AI | ✓ | ✓ | ✓ |
| OpenRouter | ✓ | ✓ | ✓ |
| OpenAI | ✓ | ✓ | ✓ |
| Microsoft Azure AIeu | ✓ | ✓ | ✓ |
| Amazon Bedrockus-east-1 | ✓ | ✗ | ✗ |
| OpenAIpriority | ✓ | ✓ | ✓ |
Tool calling: 7 of 7 listings say yes. JSON output: 6 of 7 listings say yes, 1 says no. Strict schema: 6 of 7 listings say yes, 1 says no.
Models people weigh against GPT-5.4
When we formed this view
Dates behind this page
Prices last checked 9d ago
What we do not know about this model yet
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 7 listings do not say whether they train on prompts.
Licence and identifiers
What the licence allowsWe hold no licence record for this model. Inside are the identifiers you need to pull it — its Hugging Face repo where we have one, our slug and a machine-readable card.
Licence
Commercial API terms. We hold no licence record for this model, so there is nothing to summarise here.
Identifiers
- Modality record
- text+image+file->text
- Catalogue slug
- openai-gpt-5-4