# Claude 3.5 Haiku ![Organization](https://img.shields.io/badge/Org-Anthropic-d97757) ![Context](https://img.shields.io/badge/Context-200K%20tokens-1f6feb) ![Pricing](https://img.shields.io/badge/Price-%241%20%2F%20%245%20per%201M-2da44e) ![License](https://img.shields.io/badge/License-Proprietary-d29922) ![Modality](https://img.shields.io/badge/Modality-Text%20%2B%20Vision-8957e5) ![Released](https://img.shields.io/badge/Released-Oct%202024-555) > **Quick answer:** Claude 3.5 Haiku (API: claude-3-5-haiku-20241022) is Anthropic's October 2024 fast small model. It scores 88.1% on HumanEval, 65% on MMLU-Pro, 30.1% on BigCodeBench, and 69.4% on MATH. At $1/$5 per 1M tokens with a 200K context window, it is the recommended Claude small-tier model for most tasks requiring speed at moderate cost. ## At a Glance **Where Claude 3.5 Haiku leads** - 88.1% HumanEval — strong function-level code completion - 65% MMLU-Pro — good academic knowledge - 200K-token context window - Low latency — designed for real-time applications - Vision support **Where it lags** - $1/$5 per 1M — 4-7× more expensive than GPT-4o mini ($0.15/$0.60) or Gemini 2.5 Flash ($0.15/$0.60) - 30.1% BigCodeBench — below frontier coding models - Superseded by Claude Haiku 4 for new deployments (if available) - July 2024 knowledge cutoff **Best for:** Claude-ecosystem integrations requiring fast, cost-moderate responses; vision tasks at the Haiku tier; teams that need Anthropic's safety properties without Sonnet-tier cost. ## What Claude 3.5 Haiku Is Claude 3.5 Haiku is Anthropic's small-model upgrade within the Claude 3.5 family, released October 22, 2024 alongside Claude 3.5 Sonnet v2 (Claude Sonnet 3.6). It replaced Claude 3 Haiku as Anthropic's recommended fast model, bringing Claude 3.5-quality instruction following and coding capability to the Haiku tier. Unlike Claude 3 Haiku ($0.25/$1.25), Claude 3.5 Haiku is priced at $1/$5 — higher cost reflecting substantially improved capability. Its 88.1% HumanEval and 69.4% MATH scores position it well above Claude 3 Haiku and competitive with GPT-4o mini on code and math tasks, despite being in the "small" model tier. For teams embedded in the Claude ecosystem and requiring fast responses with vision at moderate cost, Claude 3.5 Haiku is the appropriate choice. For purely cost-optimised high-volume workloads, GPT-4o mini and Gemini 2.5 Flash offer better price/performance. ## Specifications | Field | Value | |---|---| | Organization | Anthropic | | Parameters | Undisclosed | | Context window | 200,000 tokens | | API model ID | claude-3-5-haiku-20241022 | | License | Proprietary (API only) | | Release date | October 22, 2024 | | Knowledge cutoff | July 2024 | | Modality | Text + Vision (multimodal) | ## Pricing | | Input (per 1M tokens) | Output (per 1M tokens) | |---|---|---| | Anthropic API | $1.00 | $5.00 | Prompt caching available. Pricing per [Anthropic pricing page](https://www.anthropic.com/pricing). ## Context Window Claude 3.5 Haiku has a **200,000-token context window** — roughly **150 pages** of text in a single request. ## Public Benchmark Scores | Benchmark | Score | Source | Date | |---|---|---|---| | [HumanEval](/benchmarks/humaneval) | 88.1% | Benchgen evaluation | 2024-10 | | [MMLU-Pro](/benchmarks/mmlu-pro) | 65% | Benchgen evaluation | 2024-10 | | [BigCodeBench](/benchmarks/bigcodebench) | 30.1% | Benchgen evaluation | 2024-10 | | [MATH](/benchmarks/math) | 69.4% | Benchgen evaluation | 2024-10 | ## Claude 3.5 Haiku vs Alternatives | Model | HumanEval | MMLU-Pro | Context | Price (in/out per 1M) | |---|---|---|---|---| | **Claude 3.5 Haiku** | 88.1% | 65% | 200K | $1 / $5 | | [Claude Haiku 3.5](/models/anthropic/claude-haiku-3-5) | — | — | 200K | $0.80 / $4 | | [GPT-4o mini](/models/openai/gpt-4o-mini) | 87.2%* | — | 128K | $0.15 / $0.60 | | [Gemini 2.5 Flash](/models/google/gemini-2-5-flash) | — | — | 1M | $0.15 / $0.60 | *From OpenAI launch report. Claude 3.5 Haiku vs GPT-4o mini: similar HumanEval, 200K vs 128K context, at 6.7× higher input cost. Claude 3.5 Haiku is appropriate when Anthropic's training properties (safety, instruction following style) are specifically needed. ## Related Models & Benchmarks - Compare: [Claude 3 Haiku](/models/anthropic/claude-3-haiku) · [Claude Sonnet 3.6](/models/anthropic/claude-sonnet-3-6) · [GPT-4o mini](/models/openai/gpt-4o-mini) - Benchmarks: [HumanEval](/benchmarks/humaneval) · [MMLU-Pro](/benchmarks/mmlu-pro) · [BigCodeBench](/benchmarks/bigcodebench) ## Frequently Asked Questions

What is Claude 3.5 Haiku?

Claude 3.5 Haiku (claude-3-5-haiku-20241022) is Anthropic's October 2024 fast small model, scoring 88.1% HumanEval and 65% MMLU-Pro at $1/$5 per 1M tokens with 200K context.

What is Claude 3.5 Haiku's context window?

Claude 3.5 Haiku supports a 200,000-token context window.

How much does Claude 3.5 Haiku cost?

$1.00 per 1M input tokens and $5.00 per 1M output tokens, with prompt caching available.

Should I use Claude 3.5 Haiku or Claude 3 Haiku?

Claude 3.5 Haiku provides significantly better performance — 88.1% vs ~35% HumanEval — at 4× higher cost. For new deployments, Claude 3.5 Haiku is strongly recommended unless cost is the absolute constraint.

--- *Specs and scores sourced from Anthropic's official Claude 3.5 Haiku announcement (October 2024) and Benchgen evaluations. Pricing cited to the Anthropic pricing page. Last updated 2026-07-24.*

Claude 3.5 Haiku — Evaluations

Evaluations