Anthropic releases the Claude 3 model family, Claude API now available in 159 countries
Anthropic has released three models: Claude 3 Opus, Sonnet, and Haiku. Opus and Sonnet are available in the Claude product and in the Claude API, which is now generally available in 159 countries. Haiku will be available soon.
Anthropic announced the Claude 3 model family in three tiers by performance: Claude 3 Haiku (fastest), Claude 3 Sonnet, and Claude 3 Opus (most capable). Opus and Sonnet are available in the Claude product and in the Claude API, with Haiku to follow. The Claude API is now Generally Available in 159 countries.
According to Anthropic, the Claude 3 Opus model outperforms the competition on common benchmarks such as MMLU, GPQA, or GSM8K, and compared to Claude 2.1 achieves a twofold improvement in accuracy on challenging open-ended questions. According to the company, the Claude 3 Sonnet model is 2x faster than Claude 2 and Claude 2.1 while offering a higher level of capability, and Opus offers speed comparable to Claude 2 and 2.1, but with a significantly higher level of capability. All three models have vision capabilities for processing photos, charts, and technical diagrams and, according to the company, refuse to answer queries near the boundaries of safety rules less often than the previous generation. The models also improve handling of multilingual content, including Spanish, Japanese, and French.
At launch, the models offer a context window of 200 thousand tokens, but according to the company they are capable of accepting inputs of over 1 million tokens, which may be made available to select clients. On the Needle In A Haystack test, the Claude 3 Opus model achieved, according to the company, over 99% accuracy in retrieving information from a large body of text. The model family is classified at AI Safety Level 2 (ASL-2).
The rest of the source article, including further details on responsible design, was not available. See the source article for details.
Why it matters
General availability of the Claude API in 159 countries expands the range of companies and developers who can deploy the models into production without a waitlist. Three performance tiers (Haiku, Sonnet, Opus) allow choosing the ratio of speed, price, and capability according to the specific task, from live chat to complex analysis.
Two audiences, two different impacts
What this means
For individuals
Users have access to three models with different ratios of speed and capability and can choose the more suitable variant for a given task; new vision capabilities allow the model to process documents, charts, and diagrams directly.
For a business
General availability of the Claude API in 159 countries allows companies to deploy Claude 3 models into production processes (customer support, data extraction, processing knowledge bases in PDF or presentation format) without waitlist restrictions and with a choice of model based on cost and speed.
DevelopmentCheck the original
Event sources
clearly official source · 1 publisher, 0 independent. We count feeds from the same owner only once.