Overview
Claude Instant was Anthropic's fast, lightweight, lower-cost member of the legacy Claude 1–2.x family. It launched on March 14, 2023 alongside the original Claude when Anthropic first opened its API for early access, and it was positioned for low-latency, high-throughput use cases — cheap, quick replies rather than the deepest reasoning. The tier shipped in three iterations: Claude Instant 1.0 (March 14, 2023), Claude Instant 1.1 (March 31, 2023), and the final Claude Instant 1.2 (August 9, 2023).
Like its larger siblings Claude 2 and Claude 2.1, Claude Instant carried a 100,000-token context window — part of the May 2023 expansion that made Claude one of the first frontier models to read roughly 75,000 words (about a full novel) in a single prompt; Anthropic's own demo loaded the entire text of The Great Gatsby into Claude-Instant. The final Claude Instant 1.2 folded in strengths from Claude 2, improving math, coding, reasoning, safety and hallucination resistance over Claude Instant 1.1 while staying the cheapest option in the lineup.
Claude Instant is fully retired. Anthropic deprecated claude-instant-1.0, 1.1 and 1.2 on September 4, 2024 and turned them off on November 6, 2024; the model IDs no longer respond and Anthropic points users to the modern Claude Haiku models as the replacement. It is documented here as a historical reference, not a usable model.
| Released | 2023-03-14 |
|---|---|
| License | Proprietary |
| Weights | API only |
| Parameters | Undisclosed |
| Context | 100,000 tokens |
| Max output | Not separately published by Anthropic |
| Architecture | Dense decoder-only Transformer trained with unsupervised pretraining, reinforcement learning from human feedback (RLHF), and Constitutional AI (a supervised plus RL alignment phase). Parameter count was never disclosed; in Anthropic's later naming it sits on the fast, low-cost "Haiku" lineage. |
| Knowledge cutoff | Early 2023 |
| Modalities | Text |
| Status | Retired. Anthropic deprecated claude-instant-1.0/1.1/1.2 on September 4, 2024 and shut them off (retired) on November 6, 2024. Requests to these model IDs now fail; the recommended replacement is Claude Haiku. |
Benchmarks
- HumanEval (Codex, coding, pass@1)58.7%
- GSM8K (grade-school math)86.7%
- MMLU (5-shot CoT, Claude Instant 1.1)73.4%
- ARC-Challenge (5-shot, Claude Instant 1.1)85.7%
Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.
Pricing
| Input | $1.63 per million tokens per million tokens |
|---|---|
| Output | $5.51 per million tokens per million tokens |
Anthropic's published pay-as-you-go pricing for Claude Instant (100K context tier), as listed in the official November 2023 model pricing sheet. Prices applied while the model was active; it has since been retired.
Strengths
- Low latency and high throughput — built for fast, responsive replies
- Cheapest tier in the legacy Claude line: $1.63 / $5.51 per million input/output tokens
- Large-for-its-era 100,000-token context window (about 75,000 words)
- Constitutional AI / RLHF alignment for helpful, honest, harmless behaviour
- Claude Instant 1.2 reduced hallucinations and improved jailbreak resistance over 1.1
Best for
- High-volume chat and customer-support assistants where speed and cost mattered more than peak reasoning
- Summarizing and extracting quotes from long documents using the 100K context window
- Lightweight classification, moderation and routing tasks
- Multilingual question answering and text generation at scale
- Prototyping and cost-sensitive workloads that later migrated to Claude Haiku
How to access
| Provider | Model ID |
|---|---|
| Anthropic ↗ | claude-instant-1.2 |
Claude (legacy 1–2.x) — every version
The full lineage of the Claude (legacy 1–2.x) line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.
| Version | Released | Context | License |
|---|---|---|---|
| Claude 2.1current | 2023-11-21 | — | Proprietary |
| Claude 2 | 2023-07-11 | — | Proprietary |
| Claude Instant | 2023-03-14 | — | Proprietary |
| Claude 1 | 2023-03-14 | — | Proprietary |
FAQ
Is Claude Instant still available?
No. Anthropic deprecated Claude Instant (claude-instant-1.0, 1.1 and 1.2) on September 4, 2024 and retired it on November 6, 2024. The model IDs no longer respond; Anthropic recommends migrating to the modern Claude Haiku models.
What was Claude Instant good for?
It was Anthropic's fast, low-cost tier — built for low-latency, high-throughput jobs like chat assistants, summarization over its 100K-token context, classification and other cost-sensitive tasks, rather than the heaviest reasoning.
How much did Claude Instant cost?
Per Anthropic's November 2023 pricing sheet, Claude Instant cost $1.63 per million input (prompt) tokens and $5.51 per million output (completion) tokens — far cheaper than the Claude 2 / 2.1 flagship at $8.00 / $24.00 per million tokens.
What is the modern equivalent of Claude Instant?
Claude Haiku. When Anthropic retired the legacy Instant models, it named the Claude Haiku family as the recommended replacement for fast, low-cost workloads.