Claude Instant
Last reviewed
Jun 3, 2026
Sources
8 citations
Review status
Source-backed
Revision
v1 · 1,083 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 3, 2026
Sources
8 citations
Review status
Source-backed
Revision
v1 · 1,083 words
Add missing citations, update stale details, or suggest a clearer explanation.
Claude Instant was a line of fast, low-cost large language models developed by Anthropic and offered through the Anthropic API from 2023 to 2024. It served as the lightweight counterpart to the company's flagship Claude models, trading some capability for lower latency and lower price, and was aimed at high-throughput workloads such as chat, classification, content moderation, and text analysis. The line spanned three public versions: Claude Instant 1.0, 1.1, and 1.2.
Anthropic introduced Claude Instant on March 14, 2023, the same day it opened API access to its first Claude model.[1][2] From the start the company offered two variants: the full Claude model and Claude Instant, which TechCrunch described at launch as "a faster, less costly derivative."[1] Both models were built using Anthropic's constitutional AI training approach, in which model behavior is shaped against a written set of principles rather than relying solely on human feedback for every output.
The split between a larger, more capable model and a smaller, cheaper one mirrored a pattern that other developers adopted around the same period. Claude Instant was positioned for tasks where speed and cost mattered more than maximum reasoning ability, including casual dialogue, text classification, moderation, summarization, and answering questions over supplied documents.[1][2] In Anthropic's later model taxonomy, which named tiers Opus, Sonnet, and Haiku, Claude Instant is generally treated as the ancestor of the small, fast Haiku tier.
The first releases, Claude Instant 1.0 and 1.1, arrived during 2023 alongside the early Claude 1.x models and shared their general approach to dialogue and document work. Anthropic did not give Claude Instant a major upgrade in the months immediately after launch; at the time of the Claude 2 release on July 11, 2023, the company noted that Instant had not received a significant update since March.[3]
That changed with Claude Instant 1.2, which Anthropic released on August 9, 2023.[4] The company said the new version incorporated strengths from Claude 2 and showed gains in math, coding, reasoning, and safety. On the Codex coding evaluation it scored 58.7 percent, up from 52.8 percent in the previous version, and on the GSM8K grade-school math benchmark it reached 86.7 percent, compared with 80.9 percent for Claude Instant 1.1.[4] Anthropic also reported that 1.2 generated longer and more structured responses, followed formatting instructions more closely, hallucinated less, and was more resistant to jailbreaks in automated red-teaming.[4] The version handled the same kinds of tasks as before: dialogue, text analysis, summarization, document comprehension, quote extraction, and question answering.[4]
Claude Instant launched without published pricing, which Anthropic detailed later.[1] In May 2023 the company expanded the context window for both Claude and Claude Instant to 100,000 tokens, equivalent to roughly 75,000 words, allowing the model to process long documents or several hundred pages of text in a single prompt.[5] As a demonstration, Anthropic loaded the full text of The Great Gatsby into Claude Instant and had it identify a single altered line in about 22 seconds.[5]
Anthropic's published rate sheet for the period listed Claude Instant with a 100,000-token context window and standard pricing of $1.63 per million prompt tokens and $5.51 per million completion tokens, well below the $11.02 and $32.68 per million tokens charged for Claude 2 prompt and completion.[6] In December 2023 the company cut Claude Instant pricing by roughly half, to about $0.80 per million input tokens and $2.40 per million output tokens.[7]
| Version | Released | Context window | Notes |
|---|---|---|---|
| Claude Instant 1.0 | March 2023 | Up to 100K tokens (from May 2023) | Launched alongside the first Claude model[1][5] |
| Claude Instant 1.1 | 2023 | Up to 100K tokens | Codex 52.8%, GSM8K 80.9% (per 1.2 comparison)[4] |
| Claude Instant 1.2 | August 9, 2023 | 100K tokens | Drew on Claude 2; Codex 58.7%, GSM8K 86.7%[4] |
| Claude Instant pricing | Prompt / input | Completion / output |
|---|---|---|
| Standard rate (2023) | $1.63 per million tokens | $5.51 per million tokens[6] |
| After December 2023 cut | about $0.80 per million tokens | about $2.40 per million tokens[7] |
For comparison, Claude 2 carried the same context window of 100,000 tokens but was priced for "superior performance on tasks that require complex reasoning," and Anthropic kept Claude 1 available at the same price as Claude 2.[6]
Anthropic and third-party documentation pointed Claude Instant at latency-sensitive and high-volume work where the lower price per token was the deciding factor. Common applications included customer-facing chat, casual dialogue, text classification and sentiment analysis, content moderation, summarization of emails and documents, and question answering over provided material.[1][4] Because the model could ingest up to 100,000 tokens at once, it was also used for tasks over long inputs, such as scanning lengthy documents for specific information.[5] Developers reached the model through the Anthropic API using identifiers such as claude-instant-1.1 and claude-instant-1.2, and it was additionally distributed through cloud platforms including Amazon Bedrock.
Anthropic moved its fast, inexpensive tier onto the Claude 3 family in March 2024 with Claude 3 Haiku, which took over the role Claude Instant had filled. On September 4, 2024, Anthropic notified developers that the Claude 1 and Claude Instant models would be retired, and the claude-instant-1.0, claude-instant-1.1, and claude-instant-1.2 identifiers were retired on November 6, 2024.[8] After retirement, requests to those models on the Anthropic API fail, and the company's deprecation documentation directs former Claude Instant users toward its current small, fast Haiku model as the recommended replacement.[8]
The line therefore had a short production life of roughly twenty months, from the March 2023 launch to the November 2024 retirement. Its place in Anthropic's lineup, a smaller and cheaper model running beside a flagship, carried forward into the Haiku tier of the Claude 3 generation and later releases.