Reading
Free Claude Just Got Haiku 5.5. ChatGPT Free Should Worry
Haiku 5.5 is on Claude's free plan, costs about 75% less than Haiku 4.5 on the API, and tops GPT-6 Luna in Anthropic's charts. Here's the catch.
Prices last read 2026-10-18
Short answer
Yes: Haiku 5.5 is out, free users can pick it, and the API price really dropped. The "beats GPT-6 Luna" part is Anthropic grading its own homework.Anthropic released Claude Haiku 5.5 on October 7, 2026. Its Haiku page says Free, Pro, Max, Team and Enterprise users can select it on Claude.ai on the web, iOS and Android. On the API it costs $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100k tokens. Anthropic puts the average saving against Haiku 4.5 at about 75%. Anthropic's benchmark table shows Haiku 5.5 ahead of OpenAI's GPT-6 Luna on every test where both models have a score.Source: Claude Haiku 5.5 (Anthropic) · 2026-10-18; Claude Haiku (Anthropic) · 2026-10-18
What actually shipped on October 7
There are two price tiers, split at 100,000-token prompts. Above that line, input costs $0.50 and output $2.50 per million tokens. For comparison, Haiku 4.5 costs $1 for input and $5 for output. A footnote on Anthropic's page breaks the claimed saving down: 90% lower for prompts up to 100k tokens and 50% lower above that, which averages out to about 75%. The same page also admits that Haiku 5.5 uses slightly more tokens per task because of a new tokenizer. Anthropic says the model is available on all platforms, including AWS, Google Cloud and Microsoft Azure, under the model ID claude-haiku-5-5.
Source: Claude Haiku 5.5 (Anthropic) · 2026-10-18; Anthropic reveals Claude Haiku 5.5 (XDA Developers) · 2026-10-18
The launch also has a quieter subscriber perk. Anthropic says monthly API credits for the Claude Platform will roll out to Max and Team subscribers during launch week: $100 a month on Max 5x, $200 on Max 20x, and up to $500 a month on Team, pooled across users. The credits work on any model. Claude's pricing page now lists the same amounts.
Source: Claude Haiku 5.5 (Anthropic) · 2026-10-18; Claude pricing · 2026-10-18
The free plan is the real story
Claude's pricing page lists Sonnet and Haiku as included on Free, while Opus and Fable are not. Anthropic says free users can "select" Haiku 5.5, so check the model picker before you start a long chat instead of assuming it's already chosen. Free is still capped: the pricing page says every plan has limits on a rolling five-hour window, and that Pro gives at least 5x more usage per session than Free.
Source: Claude Haiku (Anthropic) · 2026-10-18; Claude pricing · 2026-10-18
Every plan has usage limits that reset on a rolling five-hour session window.
Claude pricing page, FAQ (claude.com/pricing), read 2026-10-18Source: Claude pricing · 2026-10-18
Why this matters: a better free model doesn't give free users more messages. You get a stronger model inside the same window, and when it runs out, you wait for the reset.
Haiku 5.5 vs GPT-6 Luna: whose scoreboard?
Anthropic's table shows big leads in a few places: OSWorld 2.1 (offline subset) at 72.4% vs 48.9%, and Terminal-Bench 4.0 at 39.2% vs 16.4%. The closest result is FrontierCode 1.1, at 46.4% vs 42.4%. There is no Luna score for Humanity's Last Exam. On the ChatGPT side, Notebookcheck reports that GPT-6 Luna reached regular chat for Free and Go users from October 8, after two weeks of being limited to the desktop app. The same report says OpenAI's model help page still listed GPT-5.6 Luna as the default, so it may take time to reach every account.
Source: Claude Haiku 5.5 (Anthropic) · 2026-10-18; ChatGPT Free: GPT-6 Luna kommt jetzt in den normalen Chat (Notebookcheck) · 2026-10-18
My take: the benchmark chart is the most overrated part of this launch, and the free plan is the most underrated. Vendor tables always flatter the vendor, and the gap shrinks to its narrowest on coding, which is the test most people actually care about. Still, the timing is brutal for OpenAI. ChatGPT Free got Luna into normal chat one day after Haiku 5.5 launched, after making free users open a desktop app for two weeks. Anthropic made no one wait. If you're a free user, run your own prompts through both this week. That beats any Elo score.
Who is affected and what to do
Your move
- Free or Pro user: open the model picker on web, iOS or Android and select Haiku 5.5 for quick everyday tasks.
- Developer on Haiku 4.5: re-measure token counts before trusting the 90% figure, because the new tokenizer uses slightly more tokens per task.
- Long prompts over 100k tokens: plan on the higher tier, where Anthropic says the saving is about 50%.
- Max or Team subscriber: check your Claude Platform account for the new monthly API credits.
- Comparing with ChatGPT Free: if GPT-6 Luna isn't in your picker yet, Notebookcheck says the rollout may not have reached your account.
Source: Claude Haiku 5.5 (Anthropic) · 2026-10-18; Claude Haiku (Anthropic) · 2026-10-18; ChatGPT Free: GPT-6 Luna kommt jetzt in den normalen Chat (Notebookcheck) · 2026-10-18
Tip
Hitting the free cap? The supported fixes are waiting for the five-hour reset or upgrading your own plan. Shared logins aren't a fix.
Optional: see Claude plan listings from a third-party seller (not official Anthropic access and not an endorsement). That link carries our partner code Sk52g, which is how a purchase is attributed to this site. Promotion code UPD3S.
Sources (all read 2026-10-18)