Access guide
Where to access Claude Haiku 5.5 and other AI models
Claude Haiku 5.5 is on the Claude API as claude-haiku-5-5 at $0.10 input and $0.50 output per 1M tokens, on every major cloud, in GitHub Copilot, and covered by the new monthly API credit on Claude Max and Team plans. This page answers the same question for every model that matters: where you can run it, what each route costs, and which free paths are real, across Claude, GPT-6, Kimi K3, GLM-5.3 and Gemini.
Last updated: October 9, 2026. Prices are snapshots of the provider pages listed under Sources, which change often. The OpenCode Go link is a Developers Digest referral link; you may receive a signup benefit and we may receive credits.
Short answer
For Claude, use the Claude API or your cloud's Bedrock, Google Cloud or Foundry listing; Haiku 5.5 is the cheap worker at $0.10 / $0.50, and a monthly API credit rolling out to Max and Team subscribers ($100 on Max 5x, $200 on Max 20x, up to $500 pooled on Team) can pay for it, as covered in the Max and Team API credit breakdown. Open-weights models (GLM-5.3, Kimi K3) are usually cheaper through routers and inference clouds, where hosts compete on price - start with OpenRouter or Vercel AI Gateway. Codex ships with every ChatGPT plan including Free, while Gemini CLI now needs a paid Gemini API key. Genuinely local means a smaller model from the local models hub.
The three access lanes
Official API
The vendor's own endpoint: the reference price, first access to new features, and the full context window. For closed models (Claude, GPT-6, Gemini) it is effectively the only lane, plus Bedrock, Google Cloud and Microsoft Foundry for Claude.
Routers and clouds
One key, consolidated billing and failover. Price competition between hosts only exists for open-weights models, which is why GLM-5.3 sells well under Z.ai's list price on some hosts while most Kimi K3 hosts still match Moonshot. Gateways such as OpenRouter and Vercel AI Gateway also carry closed models like Claude Haiku 5.5, where you buy convenience rather than a discount. The optionality argument is the deeper reason to care.
Local and self-host
Zero per-token cost if the weights are open and your hardware is real. The 2026 frontier open models are datacenter-class, so local for most developers means a smaller dense model via Ollama - see the local models hub for what actually fits a laptop.
Access at a glance
| Model | Official API ($/1M in / out) | Lower-cost routes | Free path | Open weights |
|---|---|---|---|---|
| Claude Haiku 5.5 | $0.10 / $0.50 up to 100K-token prompts; $0.50 / $2.50 above | Batch $0.05 / $0.25 | Max and Team monthly API credit | No |
| Claude Sonnet 5.5 | $2 / $10, cache reads $0.10 | Batch $1 / $5 | Claude plans; Max and Team API credit | No |
| Claude Opus 5.5 | $4 / $20 (fast mode $8 / $40) | Batch $2 / $10 | Claude plans; Max and Team API credit | No |
| Claude Fable 5.1 | $10 / $50, cache reads $0.25 | Batch $5 / $25 | Usage credits on Pro; 50% of weekly limits on Max | No |
| GPT-6 family (Astra / 6.1 Sol / Luna) | $10 / $50, $2 / $10, $0.10 / $0.50 | Luna $0.10 / $0.50 | Codex on ChatGPT Free and Go | No |
| Kimi K3 | $3 / $15, $0.30 cached (Moonshot) | Boundless $2.30 / $11.40 (Vercel AI Gateway) | None listed | Yes (Kimi K3 License) |
| GLM-5.3 | $1.40 / $4.40 (Z.ai) | DeepInfra $0.56 / $2.50 (Vercel AI Gateway); Flash $0.15 / $0.50 | None listed | Yes (GLM-5.2 under MIT) |
| Gemini | See Google AI pricing | Paid Gemini API key | No free CLI quota since Jun 18, 2026 | No |
Prices are per 1M input / output tokens at standard context, from each vendor's pricing page and Vercel AI Gateway's model pages on October 9, 2026. Third-party host rates are one gateway's listing; other routers list other hosts at other rates. Gemini API pricing: see Google's AI pricing page.
Model by model
Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5
AnthropicVerified October 9, 2026
Anthropic's current lineup below Fable: Opus 5.5 (September 22), Sonnet 5.5 (September 28) and Haiku 5.5 (October 7). All three run on the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS, with a 1M-token context window and 128K max output. Max and Team plans now include a monthly API credit that pays for them on the Claude Platform.
Official
- Haiku 5.5 (claude-haiku-5-5): $0.10 input / $0.50 output per 1M tokens for prompts up to 100,000 tokens, $0.50 / $2.50 above that. Cache reads $0.01 under the threshold.
- Sonnet 5.5 (claude-sonnet-5-5): $2 / $10, with cache reads cut from $0.20 to $0.10 on October 7.
- Opus 5.5 (claude-opus-5-5): $4 / $20, cache reads $0.20, fast mode $8 / $40 on the Claude API only.
- Message Batches halves every rate: Haiku 5.5 $0.05 / $0.25, Sonnet 5.5 $1 / $5, Opus 5.5 $2 / $10.
Routers and clouds
- Same model on the clouds: Bedrock ids are anthropic.claude-haiku-5-5, anthropic.claude-sonnet-5-5 and anthropic.claude-opus-5-5; Google Cloud, Foundry and Claude Platform on AWS use the Claude API ids.
- GitHub Copilot: Haiku 5.5 is generally available on Copilot Pro, Pro+, Max, Business and Enterprise, in VS Code, JetBrains, Copilot CLI and the cloud agent.
- Vercel AI Gateway lists Haiku 5.5 as anthropic/claude-haiku-5.5, with no markup on token prices.
Local and self-host
- Closed weights - no local option.
Free and near-free
- A monthly Claude API credit is rolling out from October 7: $100 on Max 5x, $200 on Max 20x, and on Team $20 per Standard seat and $100 per Premium seat, pooled and capped at $500. It covers the Claude API, Managed Agents, the Agent SDK and the Playground. New subscribers can claim it after 7 days on an eligible plan.
- The credit does not cover Claude Code, extra usage in the Claude apps, or Claude on Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS. It expires each billing cycle with no rollover. Free, Pro and Enterprise plans are not eligible.
- OpenCode Go ($10/month) lists Claude Haiku 5.5 in its model lineup.
Claude Fable 5.1
AnthropicVerified October 9, 2026
Anthropic's top generally available model, for demanding reasoning and long-horizon agentic work. Fable 5.1 replaced Fable 5, which is now listed as a legacy model, at the same $10 / $50 list price but with far cheaper cache reads.
Official
- Claude API (claude-fable-5-1): $10 input / $50 output per 1M tokens, cache reads $0.25 (Fable 5 was $1), Batch $5 / $25.
- Also on Amazon Bedrock (anthropic.claude-fable-5-1), Google Cloud, Microsoft Foundry and Claude Platform on AWS.
Routers and clouds
- Claude plans: claude.com/pricing lists Fable as usage credits on Pro, 50% of weekly limits on Max 5x and Max 20x, 50% of weekly limits on Team premium seats, and included on Enterprise.
Local and self-host
- Closed weights - no local option.
Free and near-free
- No free tier. The Max and Team monthly API credit can be spent on Fable 5.1 through the Claude API, but at $50 per 1M output tokens a $100 credit goes quickly.
GPT-6.1 Sol and the GPT-6 family
OpenAIVerified October 9, 2026
OpenAI's GPT-6 generation shipped in September in three tiers: Astra for maximum capability, Sol for coding and agentic work (now GPT-6.1 Sol), and Luna as the high-volume floor. Codex is bundled into every ChatGPT plan, which keeps it the widest zero-setup route to an OpenAI model.
Official
- API, short context: GPT-6.1 Sol $2 / $10 per 1M tokens (cached input $0.10), GPT-6 Astra $10 / $50 (cached $1), GPT-6 Luna $0.10 / $0.50 (cached $0.01). Long-context requests cost more on every tier.
- Codex is included across ChatGPT plans, including Free and Go. Codex Cloud is limited to eligible Plus, Pro, Business, Enterprise, Healthcare and Education accounts.
Routers and clouds
- Closed weights: access is the OpenAI API and ChatGPT plan bundling; there is no third-party hosting market.
Local and self-host
- Closed weights - no local option.
Free and near-free
- Codex on the ChatGPT Free plan is a real trial path for lighter local use; Codex Cloud needs a paid plan.
Kimi K3
Moonshot AIVerified October 9, 2026
Open weights under the Kimi K3 License (not MIT), served by more than a dozen providers. Most still charge Moonshot's $3 / $0.30 cached / $15 per 1M tokens, but a few hosts now undercut it, so billing model, data boundary and harness still matter more than the sticker.
Official
- Moonshot API: model kimi-k3, OpenAI-compatible. $3 input / $0.30 cached input / $15 output per 1M tokens, 1,048,576-token context.
- Kimi Code is a separate subscription covering the desktop app, CLI and IDE plugins. Its API is not interchangeable with the pay-as-you-go Kimi API.
Routers and clouds
- Vercel AI Gateway lists 17 K3 providers. Most, including Moonshot, Fireworks, Baseten, Modal and Bedrock, charge $3 / $15. Boundless lists $2.30 / $11.40, Together AI $2.70 / $13.50, and DeepInfra and DigitalOcean $2.85 / $14.25.
Local and self-host
- Weights are on Hugging Face: 2.8T total parameters with MXFP4 weights, recommended on vLLM, SGLang or TokenSpeed. Self-hosting is a datacenter project, not a local option.
Free and near-free
- The Kimi API is pay-as-you-go with no free tier listed. Vercel AI Gateway's free tier gives every team $5 a month of credits, but only on a subset of models, so check eligibility before counting on it.
- OpenCode Go ($10/month, Go Plus $40/month) lists Kimi K3 in its lineup.
GLM-5.3 and GLM-5.2
Z.aiVerified October 9, 2026
Open weights hosted almost everywhere, so several hosts undercut the first-party price. GLM-5.3 is the newer model, at the same list price as GLM-5.2, and GLM-5.3-Flash is the budget tier. Cheap hosts may serve different precisions; test your task before optimizing purely on price.
Official
- Z.ai per-token API: GLM-5.3 and GLM-5.2 both $1.40 input / $4.40 output per 1M tokens, cached input $0.26. GLM-5.3-Flash $0.15 / $0.50, GLM-5.3-FlashX $0.37 / $1.25.
- GLM Coding Plan: Lite $18, Pro $80, Max $168 at the monthly rate (lower on longer billing), usable from Claude Code and 20+ other agent tools.
Routers and clouds
- Vercel AI Gateway lists 22 GLM-5.3 providers. DeepInfra is $0.56 / $2.50 and Novita $0.91 / $2.86, while Together AI, Fireworks and Modal match Z.ai's $1.40 / $4.40.
- OpenRouter lists 33 GLM-5.3 providers. Some undercut these rates on input or output, but rarely on both, so compare against your own input-to-output mix.
Local and self-host
- GLM-5.2 weights are MIT-licensed on Hugging Face, and GLM-5.3 weights are published there too. At 753B parameters with vLLM and SGLang support, both are datacenter-class, not laptop-class.
Free and near-free
- Z.ai lists no free tier for GLM-5.3; its free models are older Flash tiers such as GLM-4.7-Flash. OpenCode Go ($10/month) includes GLM-5.3-Flash, and the referral link below may add a signup benefit.
Gemini
GoogleVerified October 9, 2026
Gemini CLI is no longer a free entry point: since June 18, 2026 it serves paid Gemini API keys, Vertex AI and Code Assist Standard or Enterprise, and unpaid-tier and Google One users were moved to Antigravity CLI. Paid API usage is priced on Google's AI pricing page.
Official
- Gemini CLI access since June 18, 2026: a paid Gemini API key, Vertex AI, or Gemini Code Assist Standard or Enterprise. Google's quota page says unpaid-tier and Google One users were moved to Antigravity CLI.
- Paid Gemini API and Vertex AI pricing: see Google's AI pricing page.
Routers and clouds
- Closed weights: the Gemini API and Vertex AI on Google's own platform are the serving paths.
Local and self-host
- Closed weights - no local option.
Free and near-free
- There is no Gemini CLI free quota any more: since June 18, 2026 the CLI needs a paid Gemini API key, Vertex AI or Code Assist Standard or Enterprise, so budget for API usage.
Also worth knowing
The Max and Team API credit
Rolling out from October 7: Max 5x includes $100 a month of Claude API credit, Max 20x $200, and Team up to $500 pooled. You claim it by linking one Console organization in claude.ai billing settings, and new subscribers wait 7 days on an eligible plan first. It pays for your own apps and agents, not Claude Code. The full rules are in the API credit breakdown.
Haiku 5.5 as the cheap worker
Haiku 5.5 bills by prompt length: $0.10 / $0.50 up to 100,000 tokens, then five times that. Keep summaries, routing and subagent calls under the threshold and it is the cheapest current Claude model by a wide margin. The Haiku 5.5 release guide has the migration steps and the tokenizer catch.
The free stack
If the budget is zero: Copilot Free (2,000 code completions a month), Codex on ChatGPT Free, Vercel AI Gateway's $5 a month of credits on eligible models, and Cline with your own key or a local Ollama model cover a surprising amount of real work.
Gateways and proxies
An Anthropic-compatible proxy can point Claude Code at NVIDIA NIM, OpenRouter, DeepSeek, LM Studio, llama.cpp, or Ollama backends. The tradeoffs - auth, logging, tool-call translation - are covered in the model gateway breakdown.
How to pick a route
Just trying a model: take the bundled or credited route - Codex on any ChatGPT plan, Vercel AI Gateway's $5 a month on eligible models, or the Claude API credit if you already pay for Max or Team. None requires a standalone API commitment.
Production tokens on a budget: for open-weights models, let the router market work for you and check what precision the cheapest hosts serve - validate on your own task first. For closed models the vendor price is the price; the levers are the tier (Haiku 5.5 or GPT-6 Luna for bounded work, Sonnet 5.5 or GPT-6.1 Sol for everyday coding), prompt caching, and the 50% Batch discount for offline jobs.
Daily agentic coding: flat-rate plans beat per-token billing once you code with a model every day: the GLM Coding Plan, Claude Pro or Max, or a ChatGPT plan with Codex. Watch where the heaviest model sits - on Max, Fable is capped at 50% of weekly limits, and on Pro it runs on usage credits, which changes which plan you need.
Control or compliance: self-host open weights on vLLM or SGLang, or pick a route with an explicit data boundary: the Claude API's US-only inference option costs 1.1x, and Bedrock and Google Cloud price their regional endpoints separately. Self-hosting only pays off at steady, high volume.
FAQ
Where can I access Claude Haiku 5.5?
On the Claude API as claude-haiku-5-5, on Amazon Bedrock as anthropic.claude-haiku-5-5, and on Google Cloud, Microsoft Foundry and Claude Platform on AWS under the Claude API id. It is also generally available in GitHub Copilot on Pro, Pro+, Max, Business and Enterprise, listed on Vercel AI Gateway as anthropic/claude-haiku-5.5, and included in OpenCode Go. List price is $0.10 input and $0.50 output per 1M tokens for prompts up to 100,000 tokens, and $0.50 / $2.50 above that.
What does the Claude Max and Team monthly API credit cover?
Rolling out from October 7, 2026: $100 a month on Max 5x, $200 on Max 20x, and $20 per Standard seat or $100 per Premium seat on Team, pooled and capped at $500. It pays for the Claude API, Claude Managed Agents, the Claude Agent SDK and the Playground in one linked Console organization. New subscribers can claim it after 7 days on an eligible plan. It does not cover Claude Code, extra usage in the Claude apps, or Claude on Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS, and unused credit expires each billing cycle.
Is Claude Fable 5.1 included in Claude subscriptions?
Partly. Anthropic's plans page lists Fable as usage credits on Pro, 50% of weekly limits on Max 5x and Max 20x and on Team premium seats, and included on Enterprise. On the API, Fable 5.1 costs $10 input / $50 output per 1M tokens, with cache reads at $0.25.
What is the cheapest way to access GLM-5.3?
There is no single cheapest host: rates change often, and many hosts trade a low input price for a high output price. Z.ai's own GLM-5.3 rate is $1.40 input / $4.40 output per 1M tokens. On Vercel AI Gateway, DeepInfra lists $0.56 / $2.50, and OpenRouter lists 33 providers at a wide spread of rates, so compare both against your input-to-output mix. GLM-5.3-Flash costs $0.15 / $0.50 direct from Z.ai if a smaller model fits the task, and for daily coding the GLM Coding Plan starts at $18 a month.
Should I use an official API, a router, or local weights?
Official APIs give the reference price and first access to new features. Routers and gateways consolidate keys and billing, add failover, and surface price competition between hosts, which matters most for open-weights models served by many providers. Local weights only make sense for open models sized for your hardware; GLM-5.3 and Kimi K3 are both datacenter-class, so most developers should use a hosted route.
Can I run any of these models locally?
GLM-5.2 has MIT-licensed weights and GLM-5.3 weights are also on Hugging Face, but at 753B parameters they target multi-GPU servers. Kimi K3's weights are open but at 2.8T parameters they are not local-viable. Claude, GPT-6 and Gemini are closed weights. For models that genuinely run on your hardware, see the local AI models hub.
Sources
- Claude API pricing (Anthropic)
- Claude models overview (Anthropic)
- Introducing Claude Haiku 5.5 (Anthropic, October 7, 2026)
- API credits for Max and Team plans (Claude Platform docs)
- Monthly API credits for Max and Team plans (Claude Help Center)
- Claude plans and pricing
- Claude Haiku 5.5 in GitHub Copilot (GitHub changelog)
- Claude Haiku 5.5 on AI Gateway (Vercel changelog)
- OpenAI API pricing
- Using Codex with your ChatGPT plan (OpenAI Help Center)
- Kimi API pricing (Moonshot AI)
- Z.ai API pricing
- Vercel AI Gateway pricing
- Kimi K3 providers on Vercel AI Gateway
- GLM-5.3 providers on Vercel AI Gateway
- GLM-5.3 providers on OpenRouter
- GLM Coding Plan (Z.ai)
- OpenCode Go
- Gemini CLI README (Google)
- Plans for GitHub Copilot (GitHub Docs)
Continue reading
- Claude Haiku 5.5: Pricing, Migration Changes, and When to Use It - the 100K pricing cliff and the migration steps
- Claude Max API Credits: $100 a Month, What They Cover - eligibility, claiming, and what the credit will not pay for
- GPT-6.1 Sol Release Guide - OpenAI's current Sol tier and its cache pricing
- Where to Run GLM-5.3 Free and Cheap - Flash, batch and provider prices
- Where to Access Kimi K3: Every Provider and Price Compared - K3 providers in detail
- Free Claude Code and the Model Gateway Tradeoffs - routing one agent across many backends
Pick the tool once you have picked the route
The tool directory has verified profiles for every router, CLI, and agent on this page, and the local models hub covers what runs on your own hardware.
Get Smarter About AI Dev
New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.