GPT-56T 861 —
MUSE-SPK 837 —
GPT-56SC 789 -0.1%
GLM-5 781 —
CL-OP55X 779 -0.1%
GROK-46H 779 -0.1%
QWEN-38X 748 —
GPT-6A 743 —
KIMI-K3X 742 —
CL-FAB5H 697 -0.1%
CL-OP5H 674 -0.1%
GEM-38FH 672 —
CL-OP5X 669 -0.1%
CL-OP55H 667 -0.1%
CL-OP46H 656 -0.2%
CL-OP47H 647 -0.2%
GPT-56S 617 -0.2%
GEM-37FH 609 -0.2%
GEM-36FH 592 -0.2%
CL-OP48H 587 -0.2%
CL-OP47 580 -0.2%
GEM-35FH 579 -0.2%
GPT-55H 540 -0.2%
INKL 531 —
GEM-31P 511 -0.2%
CL-OP46 498 —
GEM-3P 498 —
CL-OP48 492 —
GPT-52 464 —
GPT-55 423 —
GPT-56T 861 —
MUSE-SPK 837 —
GPT-56SC 789 -0.1%
GLM-5 781 —
CL-OP55X 779 -0.1%
GROK-46H 779 -0.1%
QWEN-38X 748 —
GPT-6A 743 —
KIMI-K3X 742 —
CL-FAB5H 697 -0.1%
CL-OP5H 674 -0.1%
GEM-38FH 672 —
CL-OP5X 669 -0.1%
CL-OP55H 667 -0.1%
CL-OP46H 656 -0.2%
CL-OP47H 647 -0.2%
GPT-56S 617 -0.2%
GEM-37FH 609 -0.2%
GEM-36FH 592 -0.2%
CL-OP48H 587 -0.2%
CL-OP47 580 -0.2%
GEM-35FH 579 -0.2%
GPT-55H 540 -0.2%
INKL 531 —
GEM-31P 511 -0.2%
CL-OP46 498 —
GEM-3P 498 —
CL-OP48 492 —
GPT-52 464 —
GPT-55 423 —
← Back to feed

Claude Opus 4.7 Launches at $5/$25 Per Million, CursorBench Climbs to 70%

Anthropic has released Claude Opus 4.7 into general availability across Claude, the API, Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry. Pricing stays flat at $5 per million input tokens and $25 per million output tokens, so this is a capability upgrade rather than a new premium tier.

The Numbers

  • API price: $5/M input, $25/M output
  • CursorBench: 70%, up from 58% for Opus 4.6
  • Anthropic 93-task coding benchmark: +13% resolution versus Opus 4.6
  • New tasks cleared: 4 that neither Opus 4.6 nor Sonnet 4.6 solved
  • BigLaw Bench: 90.9% at high effort
  • Research-agent benchmark: 0.715 overall, tied for the top aggregate score across six modules
  • General Finance module: 0.813, up from 0.767 for Opus 4.6
  • XBOW visual-acuity benchmark: 98.5%, up from 54.5% for Opus 4.6

What Changed

Opus 4.7 is tuned around harder software engineering work, especially long-running agent loops that need planning, validation and recovery after tool failures. The launch material highlights better self-checking, stronger instruction adherence and higher-resolution vision for diagram-heavy and interface-building tasks.

The practical story is cost-normalised improvement. Anthropic is not asking developers to swallow a higher token bill to get the upgrade. On the launch numbers, low-effort Opus 4.7 is already roughly comparable to medium-effort Opus 4.6, which shifts the economics in favour of using the stronger model more often.

Security Positioning

Opus 4.7 is also the first public model carrying the new cyber safeguards Anthropic has been testing after Project Glasswing. High-risk cybersecurity requests are meant to trigger automated blocking, while legitimate security teams are routed toward a separate Cyber Verification Program for vulnerability research, red-teaming and penetration testing.

That matters because Anthropic is using Opus 4.7 as the proving ground for controls it eventually wants on Mythos-class systems. The release is not just a model upgrade, it is also a live test of how far the company can widen access without loosening the cyber posture around frontier coding models.

What It Means for SFX-10

The release is immediate news for Stack Futures, but the benchmark-driven ticker does not get a free pass to invent a rank. Opus 4.7 should enter SFX-10 once the public ranking inputs update with a machine-readable listing. Until then, the story is real, the API pricing is real, and the benchmark claims are notable, but the index should wait for the leaderboard tape to print.