Curiosity · AI Model
Claude 2.1
Claude 2.1 shipped in November 2023 and was the first widely available LLM with a 200K-token context window. It combined better calibration on factual questions with a more helpful refusal policy than Claude 2, and served as Anthropic's default API model until Claude 3 launched in early 2024.
Model specs
- Vendor
- Anthropic
- Family
- Claude 2
- Released
- 2023-11
- Context window
- 200,000 tokens
- Modalities
- text
- Input price
- $8/M tok
- Output price
- $24/M tok
- Pricing as of
- 2024-06
Strengths
- First production-grade 200K-token context window
- More calibrated honesty / refusal behaviour than Claude 2
- Strong long-document recall at launch
Limitations
- Deprecated — superseded by Claude 3 and Claude 4 families
- No vision, tool use, or structured-output mode
- Reasoning and coding well below current Sonnet/Opus tiers
Use cases
- Legacy long-document summarisation pipelines
- Archived enterprise chatbots built on the 2023 Anthropic stack
- Historical benchmark baselines
Benchmarks
| Benchmark | Score | As of |
|---|---|---|
| MMLU | ≈78% | 2023-11 |
| HumanEval | ≈71% | 2023-11 |
Frequently asked questions
What is Claude 2.1?
Claude 2.1 is Anthropic's November 2023 flagship LLM, best known for being the first widely available model with a 200K-token context window and improved refusal calibration over Claude 2.
Is Claude 2.1 still available?
Claude 2.1 is a legacy model. Anthropic has deprecated it for new projects and recommends migrating to Claude 3.5 Sonnet or newer Claude 4 tier models.
What replaces Claude 2.1?
For most long-context workloads, Claude 3.5 Sonnet or Claude Sonnet 4 is the cheaper, smarter successor; Opus 4.7 is the top-tier option where quality dominates.
Sources
- Anthropic — Claude 2.1 announcement — accessed 2026-04-20
- Anthropic — Model deprecations — accessed 2026-04-20