---
title: "Claude Haiku 5.5: a direct rival to GPT-6 Luna, available in iolys"
description: "Haiku 5.5 takes on GPT-6 Luna for fast, affordable coding tasks. Compare prices and try it through Claude Code, OpenRouter, OpenCode Go or Zen in iolys."
language: "en"
url: "https://getiolys.com/blog/2026/10/08/claude-haiku-5-5-vs-luna-iolys"
date: "2026-10-08"
updated: "2026-10-08"
author: "J\u00E9r\u00F4me Giacomini"
tags: ["Claude Haiku 5.5","GPT-6 Luna","Claude","OpenRouter","OpenCode","Pricing","C#",".NET"]
---

# Claude Haiku 5.5: a direct rival to GPT-6 Luna, available in iolys

Anthropic released **Claude Haiku 5.5 on October 7, 2026**, for fast, frequent tasks where cost matters. It is a direct rival to GPT-6 Luna: both target focused work at scale, with the same starting API prices. [Anthropic announcement](https://www.anthropic.com/claude-haiku-5-5), [GPT-6 Luna specifications](https://developers.openai.com/api/docs/models/gpt-6-luna).

![Claude Haiku 5.5 and GPT-6 Luna: fast, affordable AI for focused tasks, available in iolys for Visual Studio 2026](https://getiolys.com/blog/media/2026/claude-haiku-5-5-vs-luna-iolys/images/haiku-luna-iolys.webp)

You can try Haiku through several providers supported by iolys, keeping your C# or .NET solution, tools, and conversations in Visual Studio 2026.

## ![Claude logo](https://getiolys.com/img/providers/claude.webp) Stronger coding at a much lower token price

For prompts up to 100,000 tokens, Haiku 5.5 has **90% lower API token prices than Haiku 4.5**. This makes repeated, focused requests easier to afford.

Standard API prices in USD per million text tokens, checked October 8, 2026:

| Token type | Haiku 4.5 (reference) | Haiku 5.5 · change vs. 4.5 |
| --- | ---: | ---: |
| Input | $1.00 | **$0.10 · ↓ −90%**{.price-change .price-change--lower} |
| Cache reads | $0.10 | **$0.01 · ↓ −90%**{.price-change .price-change--lower} |
| Output | $5.00 | **$0.50 · ↓ −90%**{.price-change .price-change--lower} |

**↓ Lower price**{.price-change .price-change--lower}. Total task costs still depend on token use and retries. [Haiku pricing](https://platform.claude.com/docs/en/about-claude/pricing).

Anthropic also reports stronger coding results. Its launch comparison puts Haiku ahead of Luna on these two evaluations:

| Coding evaluation | GPT-6 Luna (reference) | Haiku 5.5 · change vs. Luna |
| --- | ---: | ---: |
| Terminal-Bench 4.0 | 16.4% | **39.2% · ↑ +22.8 points**{.score-change .score-change--higher} |
| FrontierCode 1.1 (Main) | 42.4% | **46.4% · ↑ +4.0 points**{.score-change .score-change--higher} |

**↑ Higher score**{.score-change .score-change--higher}. Differences are percentage points, not relative percentage gains.

Terminal-Bench tests tasks in a command-line environment. FrontierCode evaluates code changes in real repositories. These are results published by Anthropic, not measurements from iolys or a guarantee for your project. [Benchmark results and methodology](https://www.anthropic.com/claude-haiku-5-5#performance).

## Haiku vs. Luna: the same starting price, different context tiers

Direct API prices in **USD per million text tokens**, checked October 8, 2026. Haiku uses its standard rates; Luna uses Standard processing. All changes below compare Haiku with Luna at the same prompt size.

### Prompts up to 100,000 tokens

| Token type | GPT-6 Luna (reference) | Haiku 5.5 · change vs. Luna |
| --- | ---: | ---: |
| Input | $0.10 | **$0.10 · = 0%**{.price-change .price-change--same} |
| Cache reads | $0.01 | **$0.01 · = 0%**{.price-change .price-change--same} |
| Output | $0.50 | **$0.50 · = 0%**{.price-change .price-change--same} |

### Longer prompts

| Prompt size | Token type | GPT-6 Luna (reference) | Haiku 5.5 · change vs. Luna |
| --- | --- | ---: | ---: |
| Over 100K to 272K | Input | $0.10 | **$0.50 · ↑ +400%**{.price-change .price-change--higher} |
| Over 100K to 272K | Cache reads | $0.01 | **$0.05 · ↑ +400%**{.price-change .price-change--higher} |
| Over 100K to 272K | Output | $0.50 | **$2.50 · ↑ +400%**{.price-change .price-change--higher} |
| Over 272K | Input | $0.20 | **$0.50 · ↑ +150%**{.price-change .price-change--higher} |
| Over 272K | Cache reads | $0.02 | **$0.05 · ↑ +150%**{.price-change .price-change--higher} |
| Over 272K | Output | $0.75 | **$2.50 · ↑ +233.3%**{.price-change .price-change--higher} |

Compared with Luna: **↓ cheaper**{.price-change .price-change--lower}, **↑ more expensive**{.price-change .price-change--higher}, **= same price**{.price-change .price-change--same}. Relative changes use Luna's price as the baseline.

Sources: [Claude API pricing](https://platform.claude.com/docs/en/about-claude/pricing), [GPT-6 Luna pricing](https://developers.openai.com/api/docs/models/gpt-6-luna).

**Long context changes the comparison.** Haiku's higher rates apply to the whole request once the prompt exceeds 100,000 tokens. Luna keeps its base rates up to 272,000 input tokens. A growing coding conversation can therefore change which model costs less. [Haiku context pricing](https://platform.claude.com/docs/en/about-claude/pricing), [Luna context pricing](https://developers.openai.com/api/docs/models/gpt-6-luna).

Haiku supports a 1-million-token context window; Luna supports 1.05 million. Capacity and price thresholds are different limits. Cache writes, paid tools, and optional processing modes can add or change charges. These API rates do not describe subscription fees or allowances. [Claude model specifications](https://platform.claude.com/docs/en/models/overview), [Luna specifications](https://developers.openai.com/api/docs/models/gpt-6-luna).

## Choose how much effort a task needs

Haiku 5.5 is the first Haiku with adjustable reasoning effort. Use less effort for simple requests and try more for tasks that need judgment. More reasoning can increase time and token use. iolys shows effort controls when the selected provider connection advertises supported levels. [Haiku effort settings](https://www.anthropic.com/claude-haiku-5-5#performance).

For complex, extended coding work, Anthropic still recommends Sonnet 5.5 or Opus 5.5. Haiku is an option for narrower tasks and supporting work. [Anthropic's model guidance](https://platform.claude.com/docs/en/models/overview).

## Available through several iolys providers

Choose the connection that fits the account you already use:

- ![Claude logo](https://getiolys.com/img/providers/claude.webp){width=24 height=24} **[Claude Code](https://getiolys.com/providers/claude)** · Use an eligible Claude subscription through the CLI managed by iolys. Haiku 5.5 is available in Claude Code; model access follows your account. [Claude availability](https://www.anthropic.com/claude/haiku).
- ![OpenRouter logo](https://getiolys.com/img/providers/openrouter.webp){width=24 height=24} **[OpenRouter](https://getiolys.com/providers/openrouter)** · Use your API key and available credit. Its catalog lists `anthropic/claude-haiku-5.5` and routes it through several upstream providers. [Haiku on OpenRouter](https://openrouter.ai/anthropic/claude-haiku-5.5).
- ![OpenCode Go logo](https://getiolys.com/img/providers/opencode-go.webp){width=24 height=24} **[OpenCode Go](https://getiolys.com/providers/opencode-go)** · Connect your Go or Go Plus API key. Haiku 5.5 is in the subscription catalog, alongside Luna, subject to your plan's allowances. [Go model catalog](https://opencode.ai/docs/go/).
- ![OpenCode Zen logo](https://getiolys.com/img/providers/opencode-zen.webp){width=24 height=24} **[OpenCode Zen](https://getiolys.com/providers/opencode-zen)** · Use your Zen API key and pay-as-you-go account. Haiku 5.5 is listed in its coding model catalog. [Zen model catalog](https://opencode.ai/docs/zen/).

Refresh your provider's model catalog or update the managed Claude Code CLI, then select Haiku 5.5 when it appears. Subscription usage and API spending remain attached to your provider account.

## What to try on your C# or .NET solution

Start with a small task whose result you can check:

- Explain a method and identify its edge cases before you change it.
- Add tests for one validation rule, then run them.
- Investigate a compiler error and propose the smallest fix.
- Summarize a diff to help you review a change.

Give Haiku and Luna the same files, instructions, and success criteria. Compare the correctness of the result, elapsed time, and total usage, including retries. That will tell you more about your workflow than token prices alone.

To compare them on your own solution, [Install iolys for Visual Studio](https://marketplace.visualstudio.com/items?itemName=iolys.iolys-visual-studio).

## Share your results on Discord

Tried Haiku 5.5 or Luna in iolys? [Join us on Discord](https://discord.gg/NMnFZmPsc5) to share your results, ask about provider setup, or suggest improvements.