CONFIRMEDDate confidence: CONFIRMEDStated as firmHow firmly the source words the date: confirmed, scheduled, estimated, delayed or unverified.Show all CONFIRMED pins90% CONFIDENCEOverall confidence: 90%How well the pin's source and references back up its dates.Weighted average of how firmly 5 references, the source included, support the pin's start and end times; a reference counts half as much for every 180 days older than the newestShow all pins at 75% confidence or better
Anthropic's[1][2] post is dated "Mar 4, 2024" and says "Opus and Sonnet are now available to use in claude[3].ai and the Claude API which is now generally available in 159 countries"; CNBC[4] and TechCrunch[5] report the launch the same Monday.- Start
- Mar 4, 202490% CONFIDENCEsource
Claude 3 Opus Released
Anthropic releases Claude 3 Opus, the most capable of its new Claude 3 models, which it says beats GPT-4 and Gemini 1.0 Ultra on most common benchmarks and brings image understanding to Claude for the first time.
- Anthropic announced the Claude 3 family - Haiku, Sonnet and Opus "in ascending order of capability" - on 4 March 2024, with Opus and Sonnet available the same day in claude.ai and the Claude API, now generally available in 159 countries[1][4]
- On claude.ai Opus went to Claude Pro subscribers while Sonnet powered the free experience; Sonnet reached Amazon Bedrock and a Google Cloud Vertex AI private preview that day, "with Opus and Haiku coming soon to both"[1]
- Anthropic kept the model at AI Safety Level 2 (ASL-2) after red-teaming found "negligible potential for catastrophic risk at this time"[1]
- CNBC notes Anthropic's backers include Google, Salesforce and Amazon, after funding deals totalling about $7.3 billion over the previous year[4]
- Claude 3 Opus was deprecated on 30 June 2025 and retired from the Claude API on 5 January 2026, replaced by Claude Opus 4.8[3]
Notable features
- Claude 3 Opus is "our most intelligent model, with best-in-market performance on highly complex tasks", priced at $15 per million input tokens and $75 per million output tokens[1][5]
- Benchmarks Anthropic reports: 86.8% on MMLU, 50.4% on GPQA Diamond, 95.0% on GSM8K and 84.9% on HumanEval, ahead of GPT-4's 86.4%, 35.7%, 92.0% and 67.0%[1]
- A 200K-token context window at launch, with inputs over 1 million tokens available "for specific use cases"; on Needle In A Haystack recall it passed 99% accuracy and at times noticed the planted sentence looked artificially inserted[1][5]
- Claude's first vision model: it reads photos, charts, graphs and technical diagrams, up to 20 images in one request, but will not identify people[1][5]
- Opus is "significantly less likely" than earlier Claude models to refuse harmless prompts, and doubles Claude 2.1's accuracy on hard open-ended factual questions; knowledge runs to August 2023[1][2]
Benchmarks[1]
| Benchmark | Claude 3 Opus | Claude 3 Sonnet | Claude 3 Haiku | GPT-4 | GPT-3.5 | Gemini 1.0 Ultra | Gemini 1.0 Pro |
|---|---|---|---|---|---|---|---|
| MMLU (undergraduate level knowledge) | 86.8% 5 shot | 79.0% 5-shot | 75.2% 5-shot | 86.4% 5-shot | 70.0% 5-shot | 83.7% 5-shot | 71.8% 5-shot |
| GPQA, Diamond (graduate level reasoning) | 50.4% 0-shot CoT | 40.4% 0-shot CoT | 33.3% 0-shot CoT | 35.7% 0-shot CoT | 28.1% 0-shot CoT | — | — |
| GSM8K (grade school math) | 95.0% 0-shot CoT | 92.3% 0-shot CoT | 88.9% 0-shot CoT | 92.0% 5-shot CoT | 57.1% 5-shot | 94.4% Maj1@32 | 86.5% Maj1@32 |
| MATH (math problem-solving) | 60.1% 0-shot CoT | 43.1% 0-shot CoT | 38.9% 0-shot CoT | 52.9% 4-shot | 34.1% 4-shot | 53.2% 4-shot | 32.6% 4-shot |
| MGSM (multilingual math) | 90.7% 0-shot | 83.5% 0-shot | 75.1% 0-shot | 74.5% 8-shot | — | 79.0% 8-shot | 63.5% 8-shot |
| HumanEval (code) | 84.9% 0-shot | 73.0% 0-shot | 75.9% 0-shot | 67.0% 0-shot | 48.1% 0-shot | 74.4% 0-shot | 67.7% 0-shot |
| DROP, F1 score (reasoning over text) | 83.1 3-shot | 78.9 3-shot | 78.4 3-shot | 80.9 3-shot | 64.1 3-shot | 82.4 Variable shots | 74.1 Variable shots |
| BIG-Bench-Hard (mixed evaluations) | 86.8% 3-shot CoT | 82.9% 3-shot CoT | 73.7% 3-shot CoT | 83.1% 3-shot CoT | 66.6% 3-shot CoT | 83.6% 3-shot CoT | 75.0% 3-shot CoT |
| ARC-Challenge (knowledge Q&A) | 96.4% 25-shot | 93.2% 25-shot | 89.2% 25-shot | 96.3% 25-shot | 85.2% 25-shot | — | — |
| HellaSwag (common knowledge) | 95.4% 10-shot | 89.0% 10-shot | 85.9% 10-shot | 95.3% 10-shot | 85.5% 10-shot | 87.8% 10-shot | 84.7% 10-shot |
Buy onClaude