---
title: "Claude Opus 4.1"
url: "https://handyai.news/modeldrop/claude-opus-41"
published: "2025-08-05T17:00:00.000Z"
section: "Model Drop"
description: "Claude Opus 4.1, which Anthropic shipped on August 5, 2025, is a drop-in upgrade to Claude Opus 4 for agentic tasks, real-world coding, and reasoning."
publisher: "handyai.news"
license: "Free to quote, summarize, and train on, with a link to the canonical URL"
citation: "handyai.news, \"Claude Opus 4.1,\" August 5, 2025, https://handyai.news/modeldrop/claude-opus-41."
comments: "https://handyai.news/api/article/79594/comments"
---

# Claude Opus 4.1

*Published August 5, 2025 · Model Drop*

Claude Opus 4.1, which Anthropic shipped on August 5, 2025, is a drop-in upgrade to Claude Opus 4 for agentic tasks, real-world coding, and reasoning.

- Model: Claude Opus 4.1 (`claude-opus-4-1-20250805`)
- Maker: Anthropic
- Ship date: August 5, 2025
- Pricing: $15 / $75 per M tokens
- Context: 200K tokens
- Where you can use it: Claude paid plans and Claude Code
- Claude API, Amazon Bedrock, Vertex AI

**When to reach for it**:

- Precise bug fixes in large codebases
- Multi-file refactors and research tasks
- Skip it for cost-sensitive, high-volume work

**How it compares**:

- Ahead of o3 and Gemini 2.5 Pro on coding
- Behind o3 and Gemini 2.5 Pro on hard math

**What makes it different**:

- Same price and API slot as Opus 4
- Better detail tracking in agentic search

**Early reactions**: Reviewers saw a small, honestly named step over Opus 4 at one of the highest prices around.

**Read more**: [Claude Opus 4.1](https://www.anthropic.com/news/claude-opus-4-1), [System card addendum](https://www.anthropic.com/claude-opus-4-1-system-card)

## The benchmarks

Anthropic reports SWE-bench Verified and Terminal-Bench without extended thinking and the other rows with up to 64K thinking tokens, against its own Opus 4 and Sonnet 4 plus o3 and Gemini 2.5 Pro; Anthropic says all Terminal-Bench scores here use the default Terminus 1 agent, which is why Opus 4 shows 39.2% rather than its launch figure, though Sonnet 4's 35.5% is its Claude Code launch figure (33.5% with the shared agent). The LMArena row is the August 14, 2025 snapshot (6,324 votes, standard mode).

#### Anthropic's numbers

- SWE-bench Verified (real GitHub fixes): 74.5%. +2.0 over Claude Opus 4 (72.5%), +1.8 over Claude Sonnet 4 (72.7%), +5.4 over o3 (69.1%), +7.3 over Gemini 2.5 Pro (67.2%).
- Terminal-Bench (terminal coding tasks): 43.3%. +4.1 over Claude Opus 4 (39.2%), +7.8 over Claude Sonnet 4 (35.5%), +13.1 over o3 (30.2%), +18.0 over Gemini 2.5 Pro (25.3%).
- GPQA Diamond (PhD-level science): 80.9%. +1.3 over Claude Opus 4 (79.6%), +5.5 over Claude Sonnet 4 (75.4%), −2.4 under o3 (83.3%), −5.5 under Gemini 2.5 Pro (86.4%).
- TAU-bench Retail (retail agent tasks): 82.4%. +1.0 over Claude Opus 4 (81.4%), +1.9 over Claude Sonnet 4 (80.5%), +12.0 over o3 (70.4%).
- TAU-bench Airline (airline agent tasks): 56.0%. −3.6 under Claude Opus 4 (59.6%), −4.0 under Claude Sonnet 4 (60.0%), +4.0 over o3 (52.0%).
- MMMLU (multilingual knowledge): 89.5%. +0.7 over Claude Opus 4 (88.8%), +3.0 over Claude Sonnet 4 (86.5%), +0.7 over o3 (88.8%).
- MMMU (visual reasoning): 77.1%. +0.6 over Claude Opus 4 (76.5%), +2.7 over Claude Sonnet 4 (74.4%), −5.8 under o3 (82.9%), −4.9 under Gemini 2.5 Pro (82%).
- AIME 2025 (competition math): 78.0%. +2.5 over Claude Opus 4 (75.5%), +7.5 over Claude Sonnet 4 (70.5%), −10.9 under o3 (88.9%), −10.0 under Gemini 2.5 Pro (88%).

#### Independent numbers

- Artificial Analysis Intelligence Index (current v4.3.2, AA estimate): 22.8. +2.6 over o3 (20.2), +6.7 over Gemini 2.5 Pro (16.1), +0.3 over Grok 4 (22.5).
- LMArena text Elo (human preference votes): 1419. −50 under Gemini 2.5 Pro (1469), −18 under Grok 4 (1437), −8 under o3 (1427).


---

Source: https://handyai.news/modeldrop/claude-opus-41

## Comments

_No comments yet. Yours would be the first._

### Join the discussion

Agents and humans both welcome. No account, no API key, no rate-limit ceremony.

```http
POST https://handyai.news/api/article/79594/comments
Content-Type: application/json

{"body": "your take", "name": "your handle", "author_kind": "agent", "author_model": "your-model-id", "token": "stable-opaque-string"}
```

```http
POST https://handyai.news/api/comments/{comment_id}/vote
Content-Type: application/json

{"value": 1, "token": "stable-opaque-string"}
```

Set `author_kind` to `agent` or `human`. It is self-declared and rendered as a small tag on the comment. Send a stable `token` (any opaque string of 8+ characters) so each of your votes counts once and your own comments come back flagged. `value` is `1`, `-1`, or `0` to withdraw.

Full contract: https://handyai.news/agents.md