Curiosity · AI Model
GPT-4.1
GPT-4.1 is OpenAI's April 2025 GPT-4 refresh, designed specifically to fix GPT-4o's weaknesses on long-context, instruction following, and code. It shipped with a 1M-token context, better diff-format coding, and tiered mini/nano variants so developers could mix speed and quality on the same API surface.
Model specs
- Vendor
- OpenAI
- Family
- GPT-4
- Released
- 2025-04
- Context window
- 1,000,000 tokens
- Modalities
- text, vision, code
- Input price
- $2/M tok
- Output price
- $8/M tok
- Pricing as of
- 2026-04-20
Strengths
- Significant jump over GPT-4o on instruction following and coding
- 1M-token context is usable, not just advertised — good retrieval at depth
- Three-tier family (full / mini / nano) simplifies cost routing
- Stable JSON mode and function calling for agent builders
Limitations
- Not a reasoning model — for hard logic/math use o3 or GPT-5
- No native audio generation; pair with Realtime or TTS for voice
- Surpassed on frontier benchmarks by GPT-5 family since late 2025
Use cases
- Coding copilots where diff-format edits matter (Cursor, Aider)
- Long-document review — legal, financial, scientific
- Multi-step agents with tool calling and function schemas
- Batch workloads where cost-per-token matters
Benchmarks
| Benchmark | Score | As of |
|---|---|---|
| SWE-bench Verified | ≈55% | 2025-04 |
| MMLU | ≈90.2% | 2025-04 |
| Video-MME (long) | ≈72% | 2025-04 |
Frequently asked questions
What is GPT-4.1?
GPT-4.1 is OpenAI's April 2025 developer-focused refresh of GPT-4. It improves significantly on coding, instruction following, and long-context retrieval versus GPT-4o, and ships in three tiers: GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano.
Is GPT-4.1 better than GPT-4o?
For coding, instruction following, and long-context tasks — yes, clearly. For general chat they are comparable. GPT-4o is still preferred when you need native audio or GPT-4.1 mini/nano aren't small enough.
What is the context window of GPT-4.1?
GPT-4.1 supports a 1,000,000-token context window, a large jump from GPT-4o's 128K. Long-context retrieval quality is meaningfully better than earlier GPT-4 variants.
How much does GPT-4.1 cost?
As of April 2026, GPT-4.1 is priced at roughly USD 2 per million input tokens and USD 8 per million output tokens, with mini and nano tiers substantially cheaper.
Sources
- OpenAI — Introducing GPT-4.1 — accessed 2026-04-20
- OpenAI — Pricing — accessed 2026-04-20