Claude Opus 4.7 Just Dropped... (Everything you need to know)

Summary
In this video, creator Productive Dude reviews Anthropic’s announcement and benchmark results for Claude Opus 4.7, released on April 16, 2026. He breaks down the model’s new capabilities, performance improvements over Opus 4.6 and competitors like GPT-5.4 and Gemini 3.1 Pro, updated features in Claude Code, and advice for managing token usage.

What is shown

• Anthropic’s blog post announcing Claude Opus 4.7, highlighting improvements in software engineering, vision, instruction following, and verification [00:00 - 00:50].
• Benchmark comparison table across Opus 4.7, Opus 4.6, GPT-5.4, Gemini 3.1 Pro, and Mythos Preview across coding, reasoning, search, tool use, computer use, and vision [00:54 - 03:57].
• Anthropic’s notes on Project Glasswing, cybersecurity safeguards, and API availability across Claude, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry [03:58 - 04:39].
• Text breakdown of new capabilities including literal instruction following, high-resolution multimodal support (up to 2,576 pixels), finance evaluations, and file-system-based memory [04:40 - 06:13].
• Bar charts for knowledge work (GDPval-AA), visual navigation (ScreenSpot-Pro), document reasoning (OfficeQA Pro), biomolecular reasoning, long-term coherence (Vending-Bench 2), and multimodal coding [06:14 - 07:51].
• Announcement details for new features: xhigh (“extra high”) effort control level, /ultrareview command in Claude Code with three free trials for Pro and Max users, and migration guidance for token budget management [07:52 - 09:16].

Claims & numbers

• Release date: The presenter states the release date is April 16, 2026 [00:01].
• Pricing: Unchanged from Opus 4.6 at $5 per million input tokens and $25 per million output tokens [04:18].
• Resolution support: Can accept images up to 2,576 pixels on the long edge (~3.75 megapixels), more than 3x prior Claude models [05:16].
• Benchmarks shown for Opus 4.7:
  • SWE-bench Pro: 64.3% (Opus 4.6: 53.4%, GPT-5.4: 57.7%, Gemini 3.1 Pro: 54.2%, Mythos Preview: 77.8%) [00:54].
  • SWE-bench Verified: 87.6% (Opus 4.6: 80.8%, Gemini 3.1 Pro: 80.6%, Mythos Preview: 93.9%) [00:54].
  • Terminal-Bench 2.0: 69.4% (Opus 4.6: 65.4%, GPT-5.4: 75.1% self-reported, Gemini 3.1 Pro: 68.5%, Mythos Preview: 82.0%) [00:54].
  • Humanity’s Last Exam: 46.9% without tools, 54.7% with tools [00:54].
  • BrowseComp (Agentic search): 79.3% (Opus 4.6: 83.7%, GPT-5.4: 89.3%) [02:04].
  • MCP-Atlas (Scaled tool use): 77.3% [02:08].
  • OSWorld Verified (Agentic computer use): 78.0% (Opus 4.6: 72.7%, GPT-5.4: 75.0%, Mythos Preview: 79.6%) [02:19].
  • Finance-Agent v1.1: 64.4% [02:52].
  • Cyber-Gym (Cybersecurity): 73.1% (Opus 4.6: 73.8%, Mythos Preview: 83.1%) [02:58].
  • GPQA Diamond: 94.2% [03:16].
  • CharXiv Reasoning (Visual reasoning): 82.1% no tools (Opus 4.6: 69.1%) [03:37].
  • MMMLU: 91.5% [03:36].
  • GDPval-AA Elo score: 1,753 (Opus 4.6: 1,619, GPT-5.4: 1,674, Gemini 3.1 Pro: 1,314) [06:19].
  • ScreenSpot-Pro (High res): 87.6% with tools, 79.5% without tools [06:22].
  • OfficeQA Pro (Document reasoning): 80.6% (Opus 4.6: 57.1%, GPT-5.4: 51.1%) [06:48].
  • Biomolecular reasoning (Structural Biology): 74.0% vs. Opus 4.6’s 30.9% [06:52].
  • Vending-Bench 2: $10,937 balance for Opus 4.7 vs. $8,018 for Opus 4.6 [07:18].
  • SWE-bench Multilingual: 80.5% vs. 77.8%; Multimodal internal: 34.5% vs. 27.1% [07:46].
• Token usage: Opus 4.7 maps to 1.0–1.35x depending on content type compared to Opus 4.6, prompting the recommendation to adjust effort settings, task budgets, or prompt conciseness [08:48].

Notable quotes

• “Opus 4.7 takes the instructions literally, and they say that users should retune their prompts and harnesses accordingly because this model is really, really good at following instructions.” [04:45]
• “Pricing remains the same as Opus 4.6, so they’re not bumping the price on this... which is good because Opus 4.6 was already expensive enough.” [04:18]
• “More than double the capability of the scoring percentage on structural biology... so based on this, this could unlock the next breakthrough in biology.” [06:59]

Assessment
This is a third-party commentary and reaction video by a tech YouTuber walking through Anthropic’s official blog announcement and benchmark graphs. The presenter does not run independent, live hands-on benchmarks in the video, relying entirely on the data and figures published in Anthropic’s release post.

Described by gemini-3.8-flash on 2026-09-29 from the video’s audio and frames.