GPT-5.4 vs GPT-5.5
Why this pair: OpenAI's consecutive prior frontier generations
Benchmark Performance
Available Benchmarks
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| ARC-AGI-1verified-v1-15fb467fd4fc · verified_score · leader | 93.6799% of row best · percent · GPT-5.4 (XHigh) | 95.00100% of row best · percent · GPT-5.5 (XHigh) |
| ARC-AGI-2verified-v2-9a3db289984e · verified_score · leader | 73.9587% of row best · percent · GPT-5.4 (XHigh) | 85.00100% of row best · percent · GPT-5.5 (XHigh) |
| LMArena Agent Arenaagent-2026-08-31-011508720696 · outcome_score · statistical tie | 3.2598% of row best · score · GPT 5.4 (High); 95% CI [2.38938480, 4.10626048]; sessions 76973; observations 3360684; rank 21 | 4.84100% of row best · score · GPT 5.5; 95% CI [3.96340788, 5.71819765]; sessions 77762; observations 1908513; rank 18 |
| LMArena Document Arenadocument-2026-07-30-011508720696 · arena_rating · statistical tie | 1,470.1499% of row best · rating · gpt-5.4; 95% CI [1463.54357191, 1476.73400825]; votes 29809; rank 15 | 1,480.30100% of row best · rating · gpt-5.5; 95% CI [1473.46488383, 1487.13183487]; votes 17286; rank 9 |
| LMArena Search Arenasearch-2026-08-24-011508720696 · arena_rating · leader | 1,197.1696% of row best · rating · gpt-5.4-search; 95% CI [1191.73811150, 1202.57682579]; votes 110116; rank 16 | 1,242.05100% of row best · rating · gpt-5.5-search; 95% CI [1236.73488085, 1247.37053078]; votes 89873; rank 3 |
| LMArena Text Arenatext-2026-09-01-011508720696 · arena_rating · leader | 1,452.5199% of row best · rating · gpt-5.4; 95% CI [1448.73439083, 1456.28324387]; votes 63615; rank 37 | 1,466.80100% of row best · rating · gpt-5.5; 95% CI [1462.99653821, 1470.60035247]; votes 64686; rank 27 |
| LMArena Vision Arenavision-2026-08-27-011508720696 · arena_rating · statistical tie | 1,293.18100% of row best · rating · gpt-5.4; 95% CI [1286.41064849, 1299.95761489]; votes 21245; rank 20 | 1,294.36100% of row best · rating · gpt-5.5-high; 95% CI [1287.65516407, 1301.05910249]; votes 20121; rank 18 |
| LiveBench2026-06-25 · overall · leader | 82.3897% of row best · percent · gpt-5.4-xhigh · 18,273 output tokens / case | 84.54100% of row best · percent · gpt-5.5-xhigh · 11,355 output tokens / case |
| ToneBench2026-08-28-10-task-cd9819ab6e4d · overall_score · leader | 82.47100% of row best · points · GPT-5.4 · 3,254 output tokens / case | 81.5399% of row best · points · GPT-5.5 (none) · 3,805 output tokens / case |
| Overall ResultCounted from the protocol-matched rows above · 3 ties | 1 benchmark win | 5 benchmark winsOverall lead |
Third-party benchmark Only like-for-like primary-publisher results are shown; raw scores, relative scores, configuration, and token spend remain visible.
Technical Differences
Side-by-Side Facts
| Field | GPT-5.4 | GPT-5.5 |
|---|---|---|
| Developer | Openai | Openai |
| Family | Gpt 5 4 | Gpt 5 5 |
| Model | GPT-5.4 | GPT-5.5 |
| Version | GPT-5.4 | GPT-5.5 |
| Lifecycle | active | active |
| Released | 2026-03-05 | 2026-04-23 |
| Knowledge cutoff | 2025-08-31 | 2025-12-01 |
| Input modalities | Unknown | Unknown |
| Output modalities | Unknown | Unknown |
| Context window | 1,050,000 | 1,050,000 |
| Total parameters | Unknown | Unknown |
| Active parameters | Unknown | Unknown |
| License | Unknown | Unknown |
| Open weights | No | No |
| API available | Yes | Yes |
| Self-hostable | No | No |
| Provider access | Openai (Standard), Openrouter (Standard) | Openai (Standard), Openrouter (Standard) |
| Capabilities | chat, generation, reasoning, structured_outputs, tools | chat, generation, reasoning, structured_outputs, tools |
13 comparable fields · 5 material differences · Interactive comparison only; indexing gate not met
GPT-5.4 Capabilities
GPT-5.5 Capabilities
Internal Comparison Graph
Related Comparisons
| A | Pair | B | Context |
|---|---|---|---|
GPT-5.3-CodexOpenAI | vs | GPT-5.4OpenAI | family variantsimage, text |
GPT-5.4 ProOpenAI | vs | GPT-5.5OpenAI | family variantsimage, text |
GPT-5.1OpenAI | vs | GPT-5.5OpenAI | family variantsimage, text |
Claude Opus 4.6Anthropic | vs | GPT-5.4OpenAI | cross-developer peersimage, text |
Claude Opus 4.6Anthropic | vs | GPT-5.5OpenAI | cross-developer peersimage, text |
Claude Opus 4.7Anthropic | vs | GPT-5.4OpenAI | cross-developer peersimage, text |
Claude Opus 4.7Anthropic | vs | GPT-5.5OpenAI | cross-developer peersimage, text |
Claude Sonnet 4.6Anthropic | vs | GPT-5.4OpenAI | cross-developer peersimage, text |
GPT-4.1OpenAI | vs | GPT-5.4OpenAI | family variantsimage, text |
GPT-4.1 MiniOpenAI | vs | GPT-5.4OpenAI | family variantsimage, text |
GPT-5.5OpenAI | vs | cross-developer peersimage, text | |
GPT-5.5OpenAI | vs | GPT-5.6 SolOpenAI | family variantsimage, text |
Primary Evidence
Sources and Freshness
Questions
GPT-5.4 vs GPT-5.5 FAQs
Is GPT-5.4 or GPT-5.5 better for coding?+
This comparison does not currently contain a protocol-matched coding benchmark for both GPT-5.4 and GPT-5.5, so Model Markets cannot name a coding leader from pricing, context size, or capability labels alone.
Which is cheaper, GPT-5.4 or GPT-5.5?+
GPT-5.4 is $2.50 and GPT-5.5 is $5.00 per million tokens, so GPT-5.4 is cheaper on this metric. GPT-5.4 is $15.00 and GPT-5.5 is $30.00 per million tokens, so GPT-5.4 is cheaper on this metric.
Which has a larger context window, GPT-5.4 or GPT-5.5?+
Neither model has a larger sourced context window in this comparison. GPT-5.4 is 1,050,000 and GPT-5.5 is 1,050,000.
Which performs better in benchmarks, GPT-5.4 or GPT-5.5?+
GPT-5.5 leads the current overall benchmark count. The result uses 9 protocol-matched benchmarks from 4 publishers; it is not a universal quality score.
Can GPT-5.4 or GPT-5.5 be self-hosted?+
Both models have the same recorded self-hosting status: unsupported. GPT-5.4 is not marked open weight; GPT-5.5 is not marked open weight.
Can GPT-5.4 and GPT-5.5 understand images?+
GPT-5.4 is not documented with image input; GPT-5.5 is not documented with image input. This reflects supported input modalities, not vision quality.
Which can generate longer answers, GPT-5.4 or GPT-5.5?+
Neither has a larger sourced maximum output. GPT-5.4 is 128,000 and GPT-5.5 is 128,000.
Do GPT-5.4 and GPT-5.5 support reasoning and tool use?+
GPT-5.4: reasoning and tool calling. GPT-5.5: reasoning and tool calling. Feature support does not establish relative quality.
Which is available from more inference providers, GPT-5.4 or GPT-5.5?+
GPT-5.4 has 2 sourced provider routes; GPT-5.5 has 2, a tie.
Which offers better value, GPT-5.4 or GPT-5.5?+
There is no universal value winner. Compare the input and output prices above with the matched benchmark result for your workload: cheaper tokens can be offset by different quality, token usage, latency, or provider availability.