GPT-4.1 vs GPT-5
At a Glance
| Compare |  GPT-4.1OpenAI | GPT-5OpenAI |
|---|---|---|
| Intelligence, Cost, and Efficiency | ||
| IntelligenceHigher is better · MM Intelligence v2.5 | #44 of 463.7 score · 2/3 sources · provisional · missing LiveBench · full-core range 2.5–35.8 | #41 of 4615.0 score · 2/3 sources · provisional · missing LiveBench · full-core range 10.0–43.3 |
| Pricing and Limits | ||
| Input priceFrom · USD / 1M tokens | $1.00Openrouter ↗ · Sep 3, 2026 | $0.625Openrouter ↗ · Sep 3, 2026 |
| Output priceFrom · USD / 1M tokens | $4.00Openrouter ↗ · Sep 3, 2026 | $5.00Openrouter ↗ · Sep 3, 2026 |
| Context windowMaximum documented tokens | 1,048K | 400K |
| Model facts checked | Aug 29, 2026View model evidence → | Sep 3, 2026View model evidence → |
Token prices are the lowest available sourced USD rates; input and output may use different providers. Cost ranking estimates output spend on LiveBench, not a full request bill. Ranking methodology →
Available Benchmarks
| Benchmark | GPT-4.1 | GPT-5 |
|---|---|---|
| ARC-AGI-1verified-v1-ba05d69f6453 · verified_score · leader | 5.508% of row best · percent · GPT-4.1 | 65.67100% of row best · percent · GPT-5 (High) |
| ARC-AGI-2verified-v2-6c676fa3e9af · verified_score · leader | 0.424% of row best · percent · GPT-4.1 | 9.86100% of row best · percent · GPT-5 (High) |
| Aider Polyglotpolyglot-225-5dc9490bb35f · pass_rate_2 · leader | 52.4460% of row best · percent · diff; aider 0.81.4.dev | 88.00100% of row best · percent · diff; aider 0.86.2.dev; reasoning high · 11,659.7 output tokens / case |
| LMArena Text Arenatext-2026-09-13-d25aabda0010 · arena_rating · leader | 1,382.7798% of row best · rating · gpt-4.1-2025-04-14; 95% CI [1379.06025569, 1386.47352582]; votes 49941; rank 157 | 1,406.06100% of row best · rating · gpt-5-high; 95% CI [1401.61960044, 1410.50585480]; votes 31428; rank 136 |
| LMArena Vision Arenavision-2026-09-13-d25aabda0010 · arena_rating · statistical tie | 1,210.03100% of row best · rating · gpt-4.1-2025-04-14; 95% CI [1203.34343280, 1216.70707848]; votes 41012; rank 74 | 1,209.22100% of row best · rating · gpt-5-high; 95% CI [1202.34954059, 1216.08580885]; votes 34351; rank 76 |
| Overall ResultCounted from the protocol-matched rows above · 1 tie | 0 benchmark wins | 4 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.
Side-by-Side Facts
| Field | GPT-4.1 | GPT-5 |
|---|---|---|
| Developer | OpenAI | OpenAI |
| Family | Gpt 4 1 | Gpt 5 |
| Model | GPT-4.1 | GPT-5 |
| Version | GPT-4.1 | GPT-5 |
| Lifecycle | active | active |
| Released | Unknown | 2025-08-07 |
| Knowledge cutoff | 2024-06-01 | 2024-09-30 |
| Input modalities | Text, Image | Text, Image |
| Output modalities | Text | Text |
| Context window | 1,048K | 400K |
| 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, tools | chat, generation, reasoning, structured_outputs, tools |
GPT-4.1 Capabilities
GPT-5 Capabilities
Related Comparisons
All image comparisons →| A | Pair | B | Context |
|---|---|---|---|
GPT-4.1OpenAI | vs | GPT-6 AstraOpenAI | family variantsimage, text |
Claude Fable 5.1Anthropic | vs | GPT-4.1OpenAI | cross-developer peersimage, text |
 Gemini 3.1 ProGoogle DeepMind | vs | GPT-4.1OpenAI | cross-developer peerstext |
 DeepSeek-V4.1-FlashDeepSeek | vs | GPT-4.1OpenAI | cross-developer peersimage, text |
GPT-4.1OpenAI | vs | Grok 4.6xAI | cross-developer peersimage, text |
GPT-4.1OpenAI | vs |  Qwen3.8-MaxQwen | cross-developer peersimage, text |
 Kimi-K3Moonshot AI | vs | GPT-4.1OpenAI | cross-developer peersimage, text |
 MiniMax-M3MiniMax | vs | GPT-4.1OpenAI | cross-developer peersimage, text |
GPT-4.1OpenAI | vs |  GLM-5.3Z.ai | cross-developer peerstext |
GPT-4.1OpenAI | vs |  Hy4 previewTencent | cross-developer peerstext |
Seed 2.1 ProByteDance Seed | vs | GPT-4.1OpenAI | cross-developer peersimage, text |
 Mistral Large 3Mistral AI | vs | GPT-4.1OpenAI | cross-developer peersimage, text |
Primary Evidence
Sources and Freshness
Questions
GPT-4.1 vs GPT-5 FAQs
Is GPT-4.1 or GPT-5 better for coding?+
GPT-5 leads the current coding subset with 1 benchmark win to 0. This describes only the matched benchmark rows shown above.
Which is cheaper, GPT-4.1 or GPT-5?+
GPT-4.1 is $1.00 and GPT-5 is $0.625 per million tokens, so GPT-5 is cheaper on this metric. GPT-4.1 is $4.00 and GPT-5 is $5.00 per million tokens, so GPT-4.1 is cheaper on this metric.
Which has a larger context window, GPT-4.1 or GPT-5?+
GPT-4.1 has the larger sourced context window. GPT-4.1 supports 1,048K and GPT-5 supports 400K.
Which performs better in benchmarks, GPT-4.1 or GPT-5?+
GPT-5 leads the current overall benchmark count. The result uses 5 protocol-matched benchmarks from 3 publishers; it is not a universal quality score.
Can GPT-4.1 or GPT-5 be self-hosted?+
Both models have the same recorded self-hosting status: unsupported. GPT-4.1 is not marked open weight; GPT-5 is not marked open weight.
Can GPT-4.1 and GPT-5 understand images?+
GPT-4.1 is documented with image input; GPT-5 is documented with image input. This reflects supported input modalities, not vision quality.
Which can generate longer answers, GPT-4.1 or GPT-5?+
GPT-5 has the larger sourced maximum output: GPT-4.1 supports 33K and GPT-5 supports 128K output tokens.
Do GPT-4.1 and GPT-5 support reasoning and tool use?+
GPT-4.1: tool calling and image input. GPT-5: reasoning, tool calling, and image input. Feature support does not establish relative quality.
Which is available from more inference providers, GPT-4.1 or GPT-5?+
GPT-4.1 has 2 sourced provider routes; GPT-5 has 2, a tie.
Which offers better value, GPT-4.1 or GPT-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.