Model Details
GPT-4 0613
Quick answer: GPT-4 0613 is the June 13, 2023 checkpoint of GPT-4, scoring 96.3% ARC-C, 95.3% HellaSwag, 86.4% MMLU, and 8.99 MT-Bench. Legacy baseline — substantially exceeded by all modern models.
At a Glance
Historical significance
- First widely-available GPT-4 version (June 2023)
- 8.99 MT-Bench — reference-quality multi-turn at Jun 2023
- 86.4% MMLU — strong general knowledge when released
Modern context
- Now superseded by GPT-4-Turbo, GPT-4.1, GPT-5, and many open-source models
- 17.6% BigCodeBench — weak complex coding by current standards
- Use only for legacy compatibility or historical benchmarking
Best for: Legacy OpenAI API compatibility; historical benchmark baselines; minimal-change production systems from 2023.
What GPT-4 0613 Is
GPT-4 0613 is OpenAI's June 13, 2023 snapshot of GPT-4. This specific checkpoint was the primary GPT-4 model through H2 2023 and is widely used in academic benchmark papers from that period.
It has been largely replaced by GPT-4-Turbo (November 2023), GPT-4.1, and the GPT-5 series. For new deployments, GPT-4.1 or GPT-5 are recommended.
Specifications
| Field | Value |
|---|---|
| Organization | OpenAI |
| License | Proprietary (API only) |
| Release date | June 13, 2023 |
| Modality | Text only |
| Context window | 8K tokens (standard), 32K tokens (GPT-4-32K variant) |
Pricing
Available via OpenAI API as a legacy endpoint. Check current pricing at platform.openai.com.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| ARC-C | 96.3% | Benchgen evaluation | 2023-06 |
| HellaSwag | 95.3% | Benchgen evaluation | 2023-06 |
| MMLU | 86.4% | Benchgen evaluation | 2023-06 |
| MT-Bench | 8.99 | Benchgen evaluation | 2023-06 |
| BigCodeBench | 17.6% | Benchgen evaluation | 2023-06 |
GPT-4 0613 vs Alternatives
| Model | MMLU | MT-Bench | BigCodeBench | Released |
|---|---|---|---|---|
| GPT-4 0613 | 86.4% | 8.99 | 17.6% | Jun 2023 |
| GPT-4.1 | — | — | — | 2025 |
| GPT-4-Turbo | — | — | — | Nov 2023 |
GPT-4 0613 is a 2023 model — all modern GPT-4.x and GPT-5 variants substantially exceed it. Use only for legacy compatibility.
Related Models & Benchmarks
- Compare: GPT-4 Turbo · GPT-4.1
- Benchmarks: MMLU · MT-Bench · ARC-C
Frequently Asked Questions
What is GPT-4 0613?
OpenAI's June 13, 2023 GPT-4 checkpoint scoring 86.4% MMLU, 8.99 MT-Bench. Legacy model — use GPT-4.1 or GPT-5 for new deployments.Specs from OpenAI's GPT-4 0613 release (June 2023) and Benchgen evaluations. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.