OpenAI · The Long-Context Stretch
GPT-4.5
The last non-reasoning OpenAI flagship. Better at vibes than reasoning. Quietly deprecated in favor of GPT-5.
By C.W. Jameson · Published 19 May 2026 · Last reviewed 19 May 2026
GPT-4.5 was OpenAI's attempt at scaling pretraining one more turn. The result was a model that wrote with palpable warmth, refused less aggressively than 4o, and lost the reasoning benchmarks to its own little sibling o3. It was deprecated within a year. Worth knowing because legacy production code still depends on it and the model's 'voice' shows up in fine-tuned derivatives.
Field signature
Notably warm prose. Lower refusal rate than 4o. Will give an opinion if asked directly.
Specifications
| Released | 2025-02 |
|---|---|
| Context window | 128,000 tokens |
| Pricing | Deprecated |
| Modalities | text · image · audio |
| License | Commercial API only, deprecated |
| Era | The Long-Context Stretch |
Strengths
- Voice and prose quality
- Conversation comfort
Weaknesses
- Reasoning benchmarks
- Officially deprecated
Authentication markers
The fingerprints by which GPT-4.5 can be identified from its output alone.
| Tell | Meaning |
|---|---|
| Em-dash use noticeably higher than GPT-4o. | GPT-4.5 lineage. |
Notable works
- Powered late-2025 ChatGPT consumer experience before GPT-5 cutover
Market position
Deprecated
Partner offer
OpenAI's API surface remains the broadest commercial offering.
Try OpenAI →Affiliate link — see disclosure.
Primary sources
- [1] OpenAI: Models
From the Almanac shop
The Operator's Compendium
Every agent harness, every routing pattern, every cost trick. 90-page PDF.
$29 — Coming soon