# Claude 3.5 Sonnet (June 2024) ![Organization](https://img.shields.io/badge/Org-Anthropic-d97757) ![Checkpoint](https://img.shields.io/badge/Checkpoint-20240620-555) ![Context](https://img.shields.io/badge/Context-200K%20tokens-1f6feb) ![Pricing](https://img.shields.io/badge/Price-%243%20%2F%20%2415%20per%201M-2da44e) ![License](https://img.shields.io/badge/License-Proprietary-d29922) ![Modality](https://img.shields.io/badge/Modality-Text%20%2B%20Vision-8957e5) ![Released](https://img.shields.io/badge/Released-Jun%202024-555) > **Quick answer:** The `claude-3-5-sonnet-20240620` checkpoint scores **49% on SWE-bench Verified**, **64% on GPQA Diamond**, and **90.7% on HumanEval** — the benchmark numbers that made it the dominant coding model of mid-2024. It runs on a 200K-token context window at $3/$15 per million tokens. The October 2024 update (`claude-3-5-sonnet-20241022`) held the same scores while adding computer use capability. ## At a Glance **Where Claude 3.5 Sonnet (June 2024) leads** - The original checkpoint that set 2024's coding benchmark standard — 49% SWE-bench Verified at Sonnet pricing - GPQA Diamond 64% — frontier-tier scientific reasoning when it launched - Stable, version-pinned checkpoint suitable for regression baselines and reproducible evaluations **Where it lags** - Superseded by `claude-3-5-sonnet-20241022` (adds computer use) and later Sonnet 4 series models - No computer use capability (added in the October 2024 update) - Knowledge cutoff April 2024 **Best for:** teams that need to reproduce evaluations or maintain a pinned baseline against the June 2024 model checkpoint specifically. ## What Claude 3.5 Sonnet (June 2024) Is The `claude-3-5-sonnet-20240620` checkpoint is the version-pinned model identifier for Anthropic's mid-2024 frontier release. Its headline benchmark numbers: **49% SWE-bench Verified** (the first sub-$20/MTok model to cross 49%), **64% GPQA Diamond** (matching frontier-tier scientific reasoning), and **90.7% HumanEval** — the highest published score at the time of release. It was the first model in the Claude 3.5 family, and set the baseline that subsequent models are measured against. Anthropic updated this checkpoint on October 22, 2024, releasing `claude-3-5-sonnet-20241022` with computer use capability and incremental improvements. The June checkpoint remains available and useful for teams that need exact reproducibility — if you ran evaluations against `20240620` and want to compare to a new model, you need the same checkpoint, not the updated version. From a Benchgen perspective, the June 2024 checkpoint is a historical reference point: it defines what "frontier-class coding at Sonnet speed" looked like at mid-2024. Teams who collected trajectory data in 2024 using this model can use it as their baseline for measuring how much the landscape has shifted since. ## Specifications | Field | Value | |---|---| | Organization | Anthropic | | API identifier | `claude-3-5-sonnet-20240620` | | Context window | 200,000 tokens | | Max output | 8,192 tokens | | License | Proprietary | | Release date | June 20, 2024 | | Knowledge cutoff | April 2024 | | Modality | Multimodal (text and vision) | | Computer use | Not supported (added in October 2024 update) | ## Pricing | | Input (per 1M tokens) | Output (per 1M tokens) | |---|---|---| | Anthropic | $3.00 | $15.00 | Pricing is identical to the October 2024 checkpoint. Prompt caching reduces input costs by up to 90%. Source: [Anthropic pricing page](https://claude.com/pricing#api). ## Context Window Claude 3.5 Sonnet (June 2024) has a **200,000-token context window** — roughly **300 pages** of text — with a maximum output of **8,192 tokens**. Specifications are identical to the October 2024 checkpoint; the update added capabilities without changing the context window or pricing. ## Public Benchmark Scores | Benchmark | Score | Source | Date | |---|---|---|---| | [SWE-bench Verified](/benchmarks/swe-bench-verified) | 49.0% | [Anthropic model card](https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf) | 2024-06 | | GPQA Diamond | 64.0% | [Anthropic announcement](https://www.anthropic.com/news/claude-3-5-sonnet) | 2024-06 | | HumanEval | 90.7% | [Anthropic announcement](https://www.anthropic.com/news/claude-3-5-sonnet) | 2024-06 | | MMLU | 88.7% | [Anthropic announcement](https://www.anthropic.com/news/claude-3-5-sonnet) | 2024-06 | *Scores are reported by Anthropic and shown for context. They are not Benchgen measurements.* ## Claude 3.5 Sonnet (June 2024) vs October 2024 Update | | June 2024 (`20240620`) | October 2024 (`20241022`) | |---|---|---| | Computer use | ✗ | ✓ | | SWE-bench Verified | 49.0% | 49.0% | | Pricing | $3 / $15 | $3 / $15 | | Context window | 200K | 200K | The core intelligence scores are unchanged between the two checkpoints; the October update's main addition is computer use. If your workflow doesn't require computer use, both checkpoints are functionally equivalent for text and vision tasks. ## How Claude 3.5 Sonnet (June 2024) Performs on Real Agent Tasks The June 2024 checkpoint's agent profile is well-documented: 64% internal agentic coding task completion at launch, strong multi-file reasoning, and reliable tool-use adherence in the instruction-tuned format. It was the first widely adopted model for Claude-based coding agents and accumulated substantial real-world trajectory data throughout 2024. For Benchgen users evaluating against historical baselines, this checkpoint is the right choice when consistency across time matters more than maximum capability. Running a Benchgen scan against `20240620` and then against the current Sonnet 4.6 is a clean way to measure how much the landscape has shifted — and to verify that your evaluation harness itself hasn't changed. ## Use Claude 3.5 Sonnet (June 2024) via API ```python from anthropic import Anthropic client = Anthropic(api_key="YOUR_API_KEY") # Pin explicitly to the June 2024 checkpoint for reproducibility message = client.messages.create( model="claude-3-5-sonnet-20240620", max_tokens=1024, messages=[{"role": "user", "content": "Review this pull request diff..."}], ) print(message.content[0].text) ``` ## Related Models & Benchmarks - Compare: [Claude 3.5 Sonnet (Oct 2024)](/models/anthropic/claude-3-5-sonnet) · [Claude Sonnet 4.6](/models/anthropic/claude-sonnet-4-6) · [GPT-4o](/models/openai/gpt-4o) - Benchmarks: [SWE-bench Verified](/benchmarks/swe-bench-verified) · [MMLU-Pro](/benchmarks/mmlu-pro) ## Frequently Asked Questions

What is the difference between claude-3-5-sonnet-20240620 and claude-3-5-sonnet-20241022?

The October 2024 checkpoint adds computer use capability (the ability to interact with desktop applications via screenshots). Intelligence benchmarks are effectively identical. If you don't need computer use, both checkpoints are functionally equivalent for text and vision tasks.

Is claude-3-5-sonnet-20240620 still available?

Yes. Anthropic keeps dated model checkpoints available for reproducibility. The June 2024 checkpoint can still be called explicitly via the API using the full `claude-3-5-sonnet-20240620` model string.

What is Claude 3.5 Sonnet (June 2024)'s context window?

200,000 tokens, with a maximum output of 8,192 tokens per response — identical to the October 2024 update.

--- *Specs and scores sourced from [Anthropic's Claude 3.5 Sonnet announcement](https://www.anthropic.com/news/claude-3-5-sonnet) and the [model card addendum](https://www-cdn.anthropic.com/fed9cc193a14b84131812372d8d5857f8f304c52/Model_Card_Claude_3_Addendum.pdf). Last updated 2026-06-19.* # Claude 3.5 Sonnet (June 2024) Claude 3.5 Sonnet (June 2024) is a large language model developed by Anthropic.

Claude 3.5 Sonnet (June 2024) — Files