GLM-52 897
GPT-56SC 873
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861
GLM-5 856
MUSE-SPK 841
QWEN-38X 824 -2.3%
GPT-6A 820
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 586 -0.5%
INKL 531
CL-OP46 497
CL-OP48 490 -0.2%
GLM-52 897
GPT-56SC 873
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861
GLM-5 856
MUSE-SPK 841
QWEN-38X 824 -2.3%
GPT-6A 820
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 586 -0.5%
INKL 531
CL-OP46 497
CL-OP48 490 -0.2%
← Back to feed

Anthropic Launches Claude Opus 5 Fast at $10/$50 Per Million — 3x Cheaper Than Opus 4.6 Fast Was

Anthropic added a fast-mode variant of Claude Opus 5 on July 24, 2026. The model is listed on OpenRouter as anthropic/claude-opus-5-fast at $10 per million input tokens and $50 per million output tokens, with cached input reads at $1/M. Anthropic describes it as “identical capabilities with higher output speed at 2x pricing relative to regular Opus 5.”

The fast mode architecture is the same underlying model as standard Opus 5 — the system that posts 96.0% SWE-bench Verified, 43.3% on Frontier-Bench v0.1 (4x the next-best GPT-5.6 Sol score of 7.8%), and 79.2% SWE-bench Pro. Fast mode trades compute efficiency for throughput, pushing tokens faster at a cost premium.

The Pricing Comparison That Matters

When Anthropic launched fast mode for Opus 4.6 in 2025, it came in at $30/$150 per million — 6x the base Opus 4.6 price. That made it expensive by any measure and restricted it to high-urgency use cases with budgets to match.

Opus 5 Fast restructures that calculus:

ModelInput/MOutput/MMultiple vs Standard
Opus 4.6 Standard$15$751x
Opus 4.6 Fast$30$1502x
Opus 5 Standard$5$251x
Opus 5 Fast$10$502x

Opus 5 Fast costs exactly one-third of what Opus 4.6 Fast cost at both tiers. For teams that were paying for Opus 4.6 Fast to get low-latency frontier capability, the upgrade to Opus 5 Fast cuts the bill by 67% while delivering a meaningfully stronger model.

The price multiple within the Opus 5 family (2x) is lower than the Opus 4.6 fast-mode multiple (2x on the base, but 6x on absolute terms vs standard). Anthropic has compressed the fast-mode premium.

Throughput and Context

Measured throughput on Anthropic’s own infrastructure is 103 tokens per second, per OpenRouter’s provider stats. Context window is 1 million tokens — same as standard Opus 5. There are no capability restrictions documented in the launch materials: all standard Opus 5 features carry over.

Cache hit pricing at $1/M for input is the same structure as standard Opus 5. For workflows with heavy prompt reuse — long-context agents, legal review pipelines, multi-turn research sessions — the effective cost after caching is substantially below the list rate.

What It Changes

Fast mode becomes relevant when latency-sensitive applications need the very best model. The prior constraint was that Opus 4.6 Fast was too expensive to run at scale for interactive use. At $10/$50, Opus 5 Fast is priced closer to what Sonnet 4.6 was charging a year ago — while delivering a generation-ahead capability level.

For coding agents running long autonomous sessions, fast mode also improves wall-clock time on multi-step tasks. Anthropic’s AA-Briefcase data puts standard Opus 5 near the top of both quality and task-completion time benchmarks; a throughput increase compounds across multi-step agent workflows.

Claude Sonnet 5 remains available at $2/$8 per million for teams where cost efficiency matters more than peak performance.

Key Numbers

  • Claude Opus 5 Fast pricing: $10/M input, $50/M output, $1/M cache read
  • Released: July 24, 2026
  • vs Opus 4.6 Fast ($30/$150): 3x cheaper at both tiers
  • vs Opus 5 Standard ($5/$25): 2x premium
  • Throughput: 103 tokens/second (Anthropic direct)
  • Context: 1M tokens
  • SWE-bench Verified: 96.0% (same underlying model as Opus 5 standard)