Feature: update the Calude AI default model to claude-haiku-4-5
## Update
### Claude AI Default Model Updated to claude-haiku-4-5
The default Anthropic Claude model used across UChat's AI features has been upgraded from the previous Haiku version to **claude-haiku-4-5** — Anthropic's fastest and most cost-efficient model, released October 2025.
#### What Changed
| | Previous Default | New Default |
|---|---|---|
| **Model ID** | `claude-haiku-3-5` | `claude-haiku-4-5` |
| **Context Window** | 200K tokens | 200K tokens |
| **Max Output** | 8,192 tokens | 64,000 tokens |
| **SWE-bench Score** | — | 73.3% (world-class coding) |
| **Pricing** | $0.80/$4 per 1M tokens | $1/$5 per 1M tokens |
| **Extended Thinking** | ✗ | ✅ (first Haiku with reasoning) |
#### What's New
- ⚡ **Faster and smarter default** — claude-haiku-4-5 matches Claude Sonnet 4's performance on coding, computer use, and agent tasks at a fraction of the cost [web:41]
- 🧠 **Extended thinking support** — the first Haiku model to support reasoning, enabling controllable reasoning depth for complex tasks [web:43]
- 📐 **8× more output tokens** — max output expanded from 8,192 to 64,000 tokens per response [web:40]
- 💰 **Cost-efficient for high-volume use** — $1/$5 per million input/output tokens, ~3× cheaper than Claude Sonnet 4.5 [web:42]
- 🔄 No action required — existing AI Agents, AI Tasks, and AI Intent Recognition will automatically use the new default model unless a specific model has been manually configured
#### Why It Matters
- All UChat workspaces on the default Anthropic/Claude provider immediately benefit from near-frontier AI performance at sub-agent speeds
- The 64K output token limit enables significantly longer, more detailed AI-generated responses in flows and live chat
- claude-haiku-4-5 is optimised for real-time, high-volume agentic applications — exactly the workload UChat AI Agents and AI Tasks generate [web:43]
> 👉 To manually select a different Claude model, go to: **AI Hub → AI Agents / AI Tasks → Edit → Provider Settings → Anthropic**
> Published: **February 20, 2026**
### Claude AI Default Model Updated to claude-haiku-4-5
The default Anthropic Claude model used across UChat's AI features has been upgraded from the previous Haiku version to **claude-haiku-4-5** — Anthropic's fastest and most cost-efficient model, released October 2025.
#### What Changed
| | Previous Default | New Default |
|---|---|---|
| **Model ID** | `claude-haiku-3-5` | `claude-haiku-4-5` |
| **Context Window** | 200K tokens | 200K tokens |
| **Max Output** | 8,192 tokens | 64,000 tokens |
| **SWE-bench Score** | — | 73.3% (world-class coding) |
| **Pricing** | $0.80/$4 per 1M tokens | $1/$5 per 1M tokens |
| **Extended Thinking** | ✗ | ✅ (first Haiku with reasoning) |
#### What's New
- ⚡ **Faster and smarter default** — claude-haiku-4-5 matches Claude Sonnet 4's performance on coding, computer use, and agent tasks at a fraction of the cost [web:41]
- 🧠 **Extended thinking support** — the first Haiku model to support reasoning, enabling controllable reasoning depth for complex tasks [web:43]
- 📐 **8× more output tokens** — max output expanded from 8,192 to 64,000 tokens per response [web:40]
- 💰 **Cost-efficient for high-volume use** — $1/$5 per million input/output tokens, ~3× cheaper than Claude Sonnet 4.5 [web:42]
- 🔄 No action required — existing AI Agents, AI Tasks, and AI Intent Recognition will automatically use the new default model unless a specific model has been manually configured
#### Why It Matters
- All UChat workspaces on the default Anthropic/Claude provider immediately benefit from near-frontier AI performance at sub-agent speeds
- The 64K output token limit enables significantly longer, more detailed AI-generated responses in flows and live chat
- claude-haiku-4-5 is optimised for real-time, high-volume agentic applications — exactly the workload UChat AI Agents and AI Tasks generate [web:43]
> 👉 To manually select a different Claude model, go to: **AI Hub → AI Agents / AI Tasks → Edit → Provider Settings → Anthropic**
> Published: **February 20, 2026**