Skip to content
Back to archive
AI Intelligence

Jul 2023

2 events · Current month →

Jul 2023

Major
Anthropic

Claude 2 — 100K Token Context Window and Claude.ai Launch

Anthropic released Claude 2, featuring a 100,000-token context window — roughly 75,000 words, or an entire novel — which far exceeded any competing model at the time. The model showed meaningful improvements over Claude 1 in coding, mathematics, and instruction following. Claude 2 was made available to the public via Claude.ai, moving Anthropic into the consumer market for the first time.

Why it mattersA 100K-token context window made entire codebases and legal documents processable in a single pass, removing the chunking-and-retrieval workaround for many document-heavy workflows.

Try itIn retrospect, the pattern to notice is that long-context models still perform better when the most relevant content is near the start or end of the prompt rather than buried in the middle.

AnthropicClaude 2Long Context100K TokensConsumer AI
Read source

Jul 2023

Landmark
Meta

LLaMA 2 Released — Meta Opens Frontier AI for Commercial Use

Meta formally released LLaMA 2 in 7B, 13B, and 70B parameter sizes under a permissive license allowing free commercial use, in partnership with Microsoft. The 70B model was competitive with GPT-3.5 on many standard benchmarks. This release established LLaMA as the de facto foundation for the open-source AI ecosystem, with thousands of derivative fine-tunes, tools, and applications building on it within months.

Why it mattersA permissive commercial license for a frontier-class open model meant teams could fine-tune and self-host without legal exposure, unlocking on-premise deployment at scale.

Try itIn retrospect, the pattern to notice is that LLaMA 2 70B fine-tuned on domain data consistently outperformed the base model on narrow tasks, validating domain-specific SFT as a practical strategy.

MetaLLaMA 2Open SourceCommercial LicenseOpen Weights
Read source