GPT-5.3 latest
GPT-5.3 model used in ChatGPT. Best general-purpose model with high intelligence and vision support. Pricing assumed same as 5.2/5.1 chat latest until announced.
About GPT-5.3 latest
GPT-5.3 Chat (latest) was OpenAI's March 2026 attempt to address some of the most persistent complaints about the GPT-5 series: too many caveats, too many dead ends, and too many wrong answers said with confidence. The headline improvements were real — 26.8% fewer hallucinations with web search and 19.7% fewer from internal knowledge compared to prior models — and users noticed a meaningfully more natural conversational rhythm. It handled everyday writing tasks well: drafting emails, refining prose, brainstorming, and summarizing documents without the hedging overload that frustrated GPT-5 adopters. That said, this model is now deprecated. OpenAI retired it on May 8, 2026, with full shutdown set for August 10, 2026. At $1.75 per million input tokens and $14.00 per million output tokens, it was also expensive for what its mid-range Intelligence Index of 54 delivered. Users who came to rely on it will need to migrate to GPT-5.5 before the deadline. For new projects, starting there is the better choice.
Best for
- Everyday conversational assistance — emails, brainstorming, quick Q&A — where reduced hedging and fewer dead ends matter
- Creative and editorial writing tasks including fiction drafting and prose refinement
- Document summarization and general information retrieval with web-augmented accuracy
- Multimodal chat workflows requiring text and image input across a 128K context window
Specs & capabilities
How GPT-5.3 latest stacks up — intelligence, speed, context, and modalities.
Intelligence
High
Speed
Medium
Context window
128,000 tokens
Max output
16,400 tokens
Knowledge cutoff
August 2025
Frequently asked questions
Is GPT-5.3 Chat still available?
No. It was deprecated on May 8, 2026, and OpenAI will shut it down completely on August 10, 2026. Existing users should migrate to GPT-5.5.
What is the context window?
128,000 tokens with a maximum output of 16,400 tokens. Note that some GPT-5.3 variants (like GPT-5.3 Instant) supported up to 400K tokens, making them better suited for large document analysis.
How much does it cost?
$1.75 per million input tokens and $14.00 per million output tokens — on the expensive side relative to its mid-range benchmark performance.
What did GPT-5.3 Chat actually improve over earlier GPT-5 models?
It cut hallucinations by roughly 19–27% depending on whether web search was used, and it reduced the excessive caveats and overly cautious phrasing that drew criticism from GPT-5 users.
Is it good for coding?
No. GPT-5.3 Chat is a general-purpose model, not optimized for software engineering. For coding tasks, GPT-5.3 Codex performs better on benchmarks like SWE-Bench.
How does it compare to GPT-5.5?
GPT-5.5 is the current successor. OpenAI moved to it quickly after GPT-5.4, and it is the recommended replacement for anyone using the gpt-5.3-chat-latest endpoint.