Compare / GPT-5.4 vs GPT-5.5

GPT-5.4 vs GPT-5.5

Which is cheaper, which is more capable, and which does better at each kind of work, from 45 results on 10 sources that measured both models.

OpenAIGPT-5.4 Released 5 Mar 2026 · $2.50 in · $15.00 out
OpenAIGPT-5.5 Released 23 Apr 2026 · $5.00 in · $30.00 out
GPT-5.4 wins 513 too close to callGPT-5.5 wins 27

Of the 45 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.

Results as of 16 November 2026 (catalogue release 2026-11-16-49bdfa045dee).

GPT-5.4 or GPT-5.5? The short answer

GPT-5.5 is the newer version; they score about the same, so pick GPT-5.4, which costs less (2.0 times less per token).

  • GPT-5.4 is cheaper: $5.62 against $11.25 per million tokens, 2.0 times less.
  • They score about the same on the Artificial Analysis Intelligence Index (39.0 and 38.4).
  • GPT-5.5 is newer: released 23 Apr 2026, against 5 Mar 2026 for GPT-5.4.
  • On third-party benchmarks, on the results both models have: GPT-5.5 does better for agents and tool use, professional work, reasoning and knowledge and writing.

Head to head

39.0
Intelligence Index
38.4
$5.62
Price per million tokens, blendedlower is better
$11.25
1,050,000
Context window, tokens
1,050,000

Every fact side by side

GPT-5.4GPT-5.5
Price per million tokens$2.50 in · $15.00 out$5.00 in · $30.00 out
Blended (3 in : 1 out)$5.62$11.25
Cheapest host, blended$5.62 (Azure)$11.25 (Azure)
Cached input, per million$0.25 cached input$0.50 cached input
Intelligence Index39.038.4
Spring Prompt overall––
Output speed, tokens a second––
Context window1,050,000 tokens1,050,000 tokens
Released5 Mar 202623 Apr 2026 (newer)
WeightsClosedClosed
Developer based inthe USthe US

Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.

Which is better for what

Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.

Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 26 of 159 27 of 159
Professional workReal tasks from banking, consulting and law, business documents and freelance projects 21 of 51 14 of 51
Reasoning and knowledgeHard questions across science, maths and general knowledge 32 of 187 26 of 187
WritingWhat people prefer in blind comparisons, and judged writing quality 21 of 169 15 of 169
Sticking to the factsSummarising without inventing things, and factual answers 37 of 139 29 of 139

From our research

Every shared result

Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.

MeasureGPT-5.4GPT-5.5
APEX-Agents reported by APEX-Agents · 1 result
Tasks passed% of tasks, higher is better 19 · 15 of 37 measured 52.4% 55.1%
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Overall · TextArena rating, higher is better 29 · 22 of 387 measured 1,475high reasoning 1,481high reasoning
Overall · Text factualityArena rating, higher is better 7 · 6 of 163 measured 1,485high reasoning 1,486high reasoning
Business, management and finance · TextArena rating, higher is better 20 · 8 of 380 measured 1,482high reasoning 1,490high reasoning
Artificial Analysis reported by Artificial Analysis · 3 results
Artificial Analysis Intelligence Indexindex score, higher is better 33 · 36 of 249 measured 39.0xhigh reasoning 38.4xhigh reasoning
GPQA Diamond% of questions, higher is better 24 · 10 of 232 measured 92.0%xhigh reasoning 93.5%xhigh reasoning
Humanity's Last Exam% of questions, higher is better 28 · 23 of 250 measured 43.7%xhigh reasoning 45.8%xhigh reasoning
Microsoft STATE-Bench reported by Microsoft STATE-Bench · 3 results
Consistency (pass^5) · STATE-Bench v0.8% of tasks, higher is better 41.8%high reasoning, statebench 44.2%high reasoning, statebench
Customer support task success (pass@1) · STATE-Bench v0.8% of tasks, higher is better 59.0%high reasoning, statebench 59.0%high reasoning, statebench
Shopping assistant task success (pass@1) · STATE-Bench v0.8% of tasks, higher is better 55.0%high reasoning, statebench 55.0%high reasoning, statebench
OpenHands Index reported by OpenHands Index · 3 results
Average% resolved, higher is better 6 · 5 of 34 measured 64.3%openhands 65.9%openhands
Front end (SWE-Bench Multimodal)% resolved, higher is better 11 · 10 of 34 measured 36.8%openhands 38.2%openhands
Greenfield (Commit0)% resolved, higher is better 3 · 8 of 34 measured 56.2%openhands 43.8%openhands
SimpleQA Verified (Epoch AI) reported by SimpleQA Verified (Epoch AI) · 1 result
Correct answers% of questions, higher is better 37 · 13 of 76 measured 45.1%xhigh reasoning 63.0%xhigh reasoning
τ²-bench reported by τ²-bench · 2 results
Consistency (pass^4) · tau2 v1.0.1, banking knowledge, alltools retrieval% of tasks, higher is better 9 · 4 of 21 measured 21.7%xhigh reasoning, tau2 29.9%xhigh reasoning, tau2
Task success (pass^1) · tau2 v1.0.1, banking knowledge, alltools retrieval% of tasks, higher is better 10 · 5 of 21 measured 39.4%xhigh reasoning, tau2 44.6%xhigh reasoning, tau2
UGI Leaderboard reported by UGI Leaderboard · 3 results
Writing score · UGIscore out of 100, higher is better 14 · 12 of 230 measured 68.8high reasoning 69.7xhigh reasoning
Requested-length error · UGI% off the requested word count, lower is better 166 · 57 of 230 measured 30.0%xhigh reasoning 13.0%xhigh reasoning
Style adherence · UGIscore from 0 to 1, higher is better 14 · 6 of 230 measured 0.40high reasoning 0.41high reasoning
Vectara Hallucination Leaderboard reported by Vectara Hallucination Leaderboard · 2 results
Hallucination rate · Vectara% of summaries, lower is better 32 · 47 of 104 measured 7.0% 9.3%
Answer rate · Vectara% of documents, higher is better 17 · 1 of 104 measured 99.9% 100.0%
Vending-Bench 2 reported by Vending-Bench 2 · 1 result
Money after a yearUS dollars, higher is better 16 · 10 of 58 measured $6,144 $7,524
Show all 45 results (23 more)
MeasureGPT-5.4GPT-5.5
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 10 results
Confirmed task success · AgentIPS effect estimate, higher is better 33 · 31 of 48 measured -0.03high reasoning, arena agent -0.01xhigh reasoning, arena agent
Creative writing · TextArena rating, higher is better 46 · 29 of 385 measured 1,442high reasoning 1,451high reasoning
Expert prompts · TextArena rating, higher is better 19 · 21 of 338 measured 1,515high reasoning 1,512high reasoning
Instruction following · TextArena rating, higher is better 30 · 19 of 387 measured 1,471high reasoning 1,478high reasoning
Overall · DocumentArena rating, higher is better 14 · 8 of 38 measured 1,470 1,483
Overall · SearchArena rating, higher is better 14 · 3 of 33 measured 1,195arena search 1,224arena search
Praise over complaint · AgentIPS effect estimate, higher is better 25 · 14 of 48 measured 0.00high reasoning, arena agent 0.08xhigh reasoning, arena agent
Steerability · AgentIPS effect estimate, higher is better 13 · 15 of 48 measured 0.05high reasoning, arena agent 0.04xhigh reasoning, arena agent
Tool grounding · AgentIPS effect estimate, higher is better 28 · 18 of 48 measured 0.00high reasoning, arena agent 0.00xhigh reasoning, arena agent
Writing, literature and language · TextArena rating, higher is better 25 · 21 of 386 measured 1,465high reasoning 1,469high reasoning
Artificial Analysis reported by Artificial Analysis · 7 results
Terminal-Bench 2.1% of tasks, higher is better 29 · 15 of 131 measured 78.3%xhigh reasoning 84.3%xhigh reasoning
τ-bench banking% of tasks, higher is better 19 · 22 of 121 measured 39.6%xhigh reasoning 39.0%xhigh reasoning
Artificial Analysis Coding Indexindex score, higher is better 26 · 13 of 130 measured 71.1xhigh reasoning 74.9xhigh reasoning
IFBench% of instructions, higher is better 31 · 21 of 195 measured 73.9%xhigh reasoning 75.9%xhigh reasoning
Long-context reasoning (AA-LCR)% of questions, higher is better 27 · 6 of 241 measured 82.0%xhigh reasoning 84.3%high reasoning
Terminal-Bench Hard% of tasks, higher is better 6 · 4 of 190 measured 57.6%xhigh reasoning 60.6%xhigh reasoning
τ²-bench telecom% of tasks, higher is better 49 · 28 of 193 measured 87.1%xhigh reasoning 93.9%xhigh reasoning
Microsoft STATE-Bench reported by Microsoft STATE-Bench · 3 results
Task success (pass@1) · STATE-Bench v0.8% of tasks, higher is better 58.6%high reasoning, statebench 58.9%high reasoning, statebench
Travel task success (pass@1) · STATE-Bench v0.8% of tasks, higher is better 62.0%high reasoning, statebench 62.0%high reasoning, statebench
User experience · STATE-Bench v0.8score from 1 to 5, higher is better 3.67high reasoning, statebench 3.62high reasoning, statebench
OpenHands Index reported by OpenHands Index · 3 results
Information gathering (GAIA)% resolved, higher is better 3 · 1 of 34 measured 82.4%openhands 86.1%openhands
Issue resolution (SWE-Bench)% resolved, higher is better 10 · 5 of 34 measured 75.6%openhands 78.2%openhands
Testing (SWT-Bench)% resolved, higher is better 12 · 3 of 34 measured 70.4%openhands 83.4%openhands

Quick answers

Is GPT-5.4 better than GPT-5.5?

GPT-5.5 is the newer version; they score about the same, so pick GPT-5.4, which costs less (2.0 times less per token).

Which is cheaper, GPT-5.4 or GPT-5.5?

GPT-5.4 is cheaper: $5.62 against $11.25 per million tokens, 2.0 times less.

Which is better for agents and tool use, GPT-5.4 or GPT-5.5?

GPT-5.4 ranks 26 of 159 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 27 for GPT-5.5.

Which is better for professional work, GPT-5.4 or GPT-5.5?

GPT-5.5 ranks 14 of 51 models on sale for real tasks from banking, consulting and law, business documents and freelance projects, against 21 for GPT-5.4.

Which is better for reasoning and knowledge, GPT-5.4 or GPT-5.5?

GPT-5.5 ranks 26 of 187 models on sale for hard questions across science, maths and general knowledge, against 32 for GPT-5.4.

Which is better for writing, GPT-5.4 or GPT-5.5?

GPT-5.5 ranks 15 of 169 models on sale for what people prefer in blind comparisons, and judged writing quality, against 21 for GPT-5.4.

Which is better for sticking to the facts, GPT-5.4 or GPT-5.5?

GPT-5.5 ranks 29 of 139 models on sale for summarising without inventing things, and factual answers, against 37 for GPT-5.4.

Run GPT-5.4 and GPT-5.5 on your own prompt.

Benchmarks aren't your task. Test both side by side on your prompt, with every answer checked against your rules and the cost of each one.