Claude Opus 5.5
Proprietary Commercial ★ Pareto-Optimal FrontierDeveloped by Anthropic · Extended Thinking Hybrid Transformer · Modalities: Text, Vision · Released 2026-09-18
What are the token costs and operational benchmarks for Claude Opus 5.5?
Verified Capability Benchmarks
Composite Intelligence: 92.7/100Architectural & Production Profile
Anthropic's September 2026 frontier powerhouse achieving 89.9% SWE-bench Verified and 92.5% GPQA Diamond at a 73% lower token price than previous Opus generations.
Top-ranked frontier intelligence leader for autonomous multi-step agentic coding, repository-scale migrations, and complex scientific reasoning at a 40% lower workload cost than Opus 5.
- Architecture
- Extended Thinking Hybrid Transformer
- Modalities
- Text, Vision
- Max Output Tokens
- 128,000 tokens
- Blended Rate (3:1)
- $8.000 / 1M
Verified Cloud Inference Endpoints Hosting Claude Opus 5.5
Compare latency, streaming speed, and token pricing across cloud API providers.
| Inference Provider | Role | Input / 1M | Output / 1M | Avg TTFT | Throughput |
|---|---|---|---|---|---|
| Anthropic Claude API → | Primary Host | $4.00 | $20.00 | 680ms | 95 tok/s |
Interactive Monthly Token Economics & ROI Forecaster
Model production workloads across prompt, completion, and cached context tokens, or switch models on the fly.
Estimated Monthly Spend
$4.00/1M in · $20.00/1M out
Prompt Cache: $0.200/1M read · $5.00/1M write
$2.00/1M in · $10.00/1M out
Prompt Cache: $0.100/1M read · $0.50/1M write
Head-to-Head Comparisons Involving Claude Opus 5.5 (36)
Claude Opus 5.5 vs Gemini 4 Argon
Gemini 4 Argon is 50.0% cheaper for input tokens ($2.00 vs. $4.00 per 1M tokens) and $10.00 vs. $20.00 for output tokens (2.0x cost advantage). In terms of operational performance, Claude Opus 5.5 delivers lower response latency at 680ms TTFT (170ms faster than Gemini 4 Argon) with 95 vs. 115 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 84.5% for Gemini 4 Argon and 93.2% vs. 93.8% on GPQA Diamond.
Claude Opus 5.5 vs Claude 4.5 Sonnet
Claude 4.5 Sonnet is 25.0% cheaper for input tokens ($3.00 vs. $4.00 per 1M tokens) and $15.00 vs. $20.00 for output tokens (1.3x cost advantage). In terms of operational performance, Claude 4.5 Sonnet delivers lower response latency at 380ms TTFT (300ms faster than Claude Opus 5.5) with 84 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 75.8% for Claude 4.5 Sonnet and 93.2% vs. 84.2% on GPQA Diamond.
Claude Opus 5.5 vs Gemini 3.8 Flash
Gemini 3.8 Flash is 93.8% cheaper for input tokens ($0.25 vs. $4.00 per 1M tokens) and $1.50 vs. $20.00 for output tokens (16.0x cost advantage). In terms of operational performance, Gemini 3.8 Flash delivers lower response latency at 105ms TTFT (575ms faster than Claude Opus 5.5) with 210 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 68.5% for Gemini 3.8 Flash and 93.2% vs. 79.2% on GPQA Diamond.
Claude Opus 5.5 vs GPT-5
GPT-5 is 37.5% cheaper for input tokens ($2.50 vs. $4.00 per 1M tokens) and $10.00 vs. $20.00 for output tokens (1.6x cost advantage). In terms of operational performance, GPT-5 delivers lower response latency at 340ms TTFT (340ms faster than Claude Opus 5.5) with 105 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 74.9% for GPT-5 and 93.2% vs. 85.4% on GPQA Diamond.
Claude Opus 5.5 vs GPT-6 Astra
Claude Opus 5.5 is 60.0% cheaper for input tokens ($4.00 vs. $10.00 per 1M tokens) and $20.00 vs. $50.00 for output tokens (2.5x cost advantage). In terms of operational performance, Claude Opus 5.5 delivers lower response latency at 680ms TTFT (60ms faster than GPT-6 Astra) with 95 vs. 74 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 75.7% for GPT-6 Astra and 93.2% vs. 96.0% on GPQA Diamond.
Claude Opus 5.5 vs DeepSeek V4 Pro
DeepSeek V4 Pro is 67.0% cheaper for input tokens ($1.32 vs. $4.00 per 1M tokens) and $3.96 vs. $20.00 for output tokens (3.0x cost advantage). In terms of operational performance, DeepSeek V4 Pro delivers lower response latency at 520ms TTFT (160ms faster than Claude Opus 5.5) with 78 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 80.6% for DeepSeek V4 Pro and 93.2% vs. 90.1% on GPQA Diamond.
Claude Opus 5.5 vs DeepSeek R1
DeepSeek R1 is 86.2% cheaper for input tokens ($0.55 vs. $4.00 per 1M tokens) and $2.19 vs. $20.00 for output tokens (7.3x cost advantage). In terms of operational performance, Claude Opus 5.5 delivers lower response latency at 680ms TTFT (770ms faster than DeepSeek R1) with 95 vs. 58 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 49.2% for DeepSeek R1 and 93.2% vs. 71.5% on GPQA Diamond.
Claude Opus 5.5 vs DeepSeek V4 Flash Vision (Exp)
DeepSeek V4 Flash Vision (Exp) is 94.5% cheaper for input tokens ($0.22 vs. $4.00 per 1M tokens) and $0.65 vs. $20.00 for output tokens (18.2x cost advantage). In terms of operational performance, DeepSeek V4 Flash Vision (Exp) delivers lower response latency at 310ms TTFT (370ms faster than Claude Opus 5.5) with 165 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 77.4% for DeepSeek V4 Flash Vision (Exp) and 93.2% vs. 91.3% on GPQA Diamond.
Claude Opus 5.5 vs OpenAI o3
OpenAI o3 is 50.0% cheaper for input tokens ($2.00 vs. $4.00 per 1M tokens) and $8.00 vs. $20.00 for output tokens (2.0x cost advantage). In terms of operational performance, Claude Opus 5.5 delivers lower response latency at 680ms TTFT (470ms faster than OpenAI o3) with 95 vs. 82 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 71.7% for OpenAI o3 and 93.2% vs. 87.7% on GPQA Diamond.
Claude Opus 5.5 vs Claude 4 Opus
Claude Opus 5.5 is 73.3% cheaper for input tokens ($4.00 vs. $15.00 per 1M tokens) and $20.00 vs. $75.00 for output tokens (3.8x cost advantage). In terms of operational performance, Claude Opus 5.5 delivers lower response latency at 680ms TTFT (240ms faster than Claude 4 Opus) with 95 vs. 46 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 77.2% for Claude 4 Opus and 93.2% vs. 87.1% on GPQA Diamond.
Claude Opus 5.5 vs Claude 3.7 Sonnet
Claude 3.7 Sonnet is 25.0% cheaper for input tokens ($3.00 vs. $4.00 per 1M tokens) and $15.00 vs. $20.00 for output tokens (1.3x cost advantage). In terms of operational performance, Claude 3.7 Sonnet delivers lower response latency at 540ms TTFT (140ms faster than Claude Opus 5.5) with 72 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 70.3% for Claude 3.7 Sonnet and 93.2% vs. 78.5% on GPQA Diamond.
Claude Opus 5.5 vs Gemini 3.1 Pro
Gemini 3.1 Pro is 68.8% cheaper for input tokens ($1.25 vs. $4.00 per 1M tokens) and $10.00 vs. $20.00 for output tokens (3.2x cost advantage). In terms of operational performance, Gemini 3.1 Pro delivers lower response latency at 420ms TTFT (260ms faster than Claude Opus 5.5) with 98 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 71.8% for Gemini 3.1 Pro and 93.2% vs. 82.4% on GPQA Diamond.
Claude Opus 5.5 vs Grok 4
Grok 4 is 37.5% cheaper for input tokens ($2.50 vs. $4.00 per 1M tokens) and $12.50 vs. $20.00 for output tokens (1.6x cost advantage). In terms of operational performance, Grok 4 delivers lower response latency at 480ms TTFT (200ms faster than Claude Opus 5.5) with 95 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 73.2% for Grok 4 and 93.2% vs. 86.2% on GPQA Diamond.
Claude Opus 5.5 vs Llama 4 Maverick
Llama 4 Maverick is 95.0% cheaper for input tokens ($0.20 vs. $4.00 per 1M tokens) and $0.60 vs. $20.00 for output tokens (20.0x cost advantage). In terms of operational performance, Llama 4 Maverick delivers lower response latency at 135ms TTFT (545ms faster than Claude Opus 5.5) with 175 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 62.4% for Llama 4 Maverick and 93.2% vs. 73.8% on GPQA Diamond.
Claude Opus 5.5 vs Qwen 3.8 Flash Next
Qwen 3.8 Flash Next is 96.0% cheaper for input tokens ($0.16 vs. $4.00 per 1M tokens) and $0.47 vs. $20.00 for output tokens (25.0x cost advantage). In terms of operational performance, Qwen 3.8 Flash Next delivers lower response latency at 280ms TTFT (400ms faster than Claude Opus 5.5) with 185 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 81.0% for Qwen 3.8 Flash Next and 93.2% vs. 91.7% on GPQA Diamond.
Claude Opus 5.5 vs Gemini 3.5 Flash-Lite
Gemini 3.5 Flash-Lite is 98.1% cheaper for input tokens ($0.07 vs. $4.00 per 1M tokens) and $0.30 vs. $20.00 for output tokens (53.3x cost advantage). In terms of operational performance, Gemini 3.5 Flash-Lite delivers lower response latency at 78ms TTFT (602ms faster than Claude Opus 5.5) with 265 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 55.4% for Gemini 3.5 Flash-Lite and 93.2% vs. 68.4% on GPQA Diamond.
Claude Opus 5.5 vs OpenAI o4-mini
OpenAI o4-mini is 72.5% cheaper for input tokens ($1.10 vs. $4.00 per 1M tokens) and $4.40 vs. $20.00 for output tokens (3.6x cost advantage). In terms of operational performance, Both models demonstrate comparable response latency around 680ms TTFT (95 vs. 125 tok/s). Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 68.1% for OpenAI o4-mini and 93.2% vs. 81.4% on GPQA Diamond.
Claude Opus 5.5 vs DeepSeek V3
DeepSeek V3 is 96.5% cheaper for input tokens ($0.14 vs. $4.00 per 1M tokens) and $0.28 vs. $20.00 for output tokens (28.6x cost advantage). In terms of operational performance, DeepSeek V3 delivers lower response latency at 410ms TTFT (270ms faster than Claude Opus 5.5) with 78 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 42.0% for DeepSeek V3 and 93.2% vs. 59.1% on GPQA Diamond.
Claude Opus 5.5 vs Gemma 4 31B IT
Gemma 4 31B IT is 97.0% cheaper for input tokens ($0.12 vs. $4.00 per 1M tokens) and $0.36 vs. $20.00 for output tokens (33.3x cost advantage). In terms of operational performance, Gemma 4 31B IT delivers lower response latency at 115ms TTFT (565ms faster than Claude Opus 5.5) with 155 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 59.2% for Gemma 4 31B IT and 93.2% vs. 71.2% on GPQA Diamond.
Claude Opus 5.5 vs OpenAI o3-mini
OpenAI o3-mini is 72.5% cheaper for input tokens ($1.10 vs. $4.00 per 1M tokens) and $4.40 vs. $20.00 for output tokens (3.6x cost advantage). In terms of operational performance, Claude Opus 5.5 delivers lower response latency at 680ms TTFT (270ms faster than OpenAI o3-mini) with 95 vs. 98 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 49.3% for OpenAI o3-mini and 93.2% vs. 79.7% on GPQA Diamond.
Claude Opus 5.5 vs Claude 3.5 Haiku
Claude 3.5 Haiku is 80.0% cheaper for input tokens ($0.80 vs. $4.00 per 1M tokens) and $4.00 vs. $20.00 for output tokens (5.0x cost advantage). In terms of operational performance, Claude 3.5 Haiku delivers lower response latency at 190ms TTFT (490ms faster than Claude Opus 5.5) with 130 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 40.6% for Claude 3.5 Haiku and 93.2% vs. 41.6% on GPQA Diamond.
Claude Opus 5.5 vs Claude 3.5 Sonnet
Claude 3.5 Sonnet is 25.0% cheaper for input tokens ($3.00 vs. $4.00 per 1M tokens) and $15.00 vs. $20.00 for output tokens (1.3x cost advantage). In terms of operational performance, Claude 3.5 Sonnet delivers lower response latency at 310ms TTFT (370ms faster than Claude Opus 5.5) with 80 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 49.0% for Claude 3.5 Sonnet and 93.2% vs. 59.4% on GPQA Diamond.
Claude Opus 5.5 vs Gemini 2.0 Flash
Gemini 2.0 Flash is 97.5% cheaper for input tokens ($0.10 vs. $4.00 per 1M tokens) and $0.40 vs. $20.00 for output tokens (40.0x cost advantage). In terms of operational performance, Gemini 2.0 Flash delivers lower response latency at 280ms TTFT (400ms faster than Claude Opus 5.5) with 185 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 51.8% for Gemini 2.0 Flash and 93.2% vs. 63.6% on GPQA Diamond.
Claude Opus 5.5 vs Gemini 2.5 Flash
Gemini 2.5 Flash is 92.5% cheaper for input tokens ($0.30 vs. $4.00 per 1M tokens) and $2.50 vs. $20.00 for output tokens (13.3x cost advantage). In terms of operational performance, Gemini 2.5 Flash delivers lower response latency at 110ms TTFT (570ms faster than Claude Opus 5.5) with 180 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 58.0% for Gemini 2.5 Flash and 93.2% vs. 70.5% on GPQA Diamond.
Claude Opus 5.5 vs Gemini 2.5 Pro
Gemini 2.5 Pro is 68.8% cheaper for input tokens ($1.25 vs. $4.00 per 1M tokens) and $10.00 vs. $20.00 for output tokens (3.2x cost advantage). In terms of operational performance, Claude Opus 5.5 delivers lower response latency at 680ms TTFT (20ms faster than Gemini 2.5 Pro) with 95 vs. 80 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 63.8% for Gemini 2.5 Pro and 93.2% vs. 74.0% on GPQA Diamond.
Claude Opus 5.5 vs GLM-5.3
GLM-5.3 is 91.2% cheaper for input tokens ($0.35 vs. $4.00 per 1M tokens) and $1.00 vs. $20.00 for output tokens (11.4x cost advantage). In terms of operational performance, GLM-5.3 delivers lower response latency at 160ms TTFT (520ms faster than Claude Opus 5.5) with 120 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 56.0% for GLM-5.3 and 93.2% vs. 67.0% on GPQA Diamond.
Claude Opus 5.5 vs GLM-5.3-Flash
GLM-5.3-Flash is 97.0% cheaper for input tokens ($0.12 vs. $4.00 per 1M tokens) and $0.40 vs. $20.00 for output tokens (33.3x cost advantage). In terms of operational performance, GLM-5.3-Flash delivers lower response latency at 90ms TTFT (590ms faster than Claude Opus 5.5) with 175 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 52.8% for GLM-5.3-Flash and 93.2% vs. 63.5% on GPQA Diamond.
Claude Opus 5.5 vs GPT-4o
GPT-4o is 37.5% cheaper for input tokens ($2.50 vs. $4.00 per 1M tokens) and $10.00 vs. $20.00 for output tokens (1.6x cost advantage). In terms of operational performance, GPT-4o delivers lower response latency at 280ms TTFT (400ms faster than Claude Opus 5.5) with 100 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 48.0% for GPT-4o and 93.2% vs. 53.6% on GPQA Diamond.
Claude Opus 5.5 vs GPT-4o Mini
GPT-4o Mini is 96.2% cheaper for input tokens ($0.15 vs. $4.00 per 1M tokens) and $0.60 vs. $20.00 for output tokens (26.7x cost advantage). In terms of operational performance, GPT-4o Mini delivers lower response latency at 220ms TTFT (460ms faster than Claude Opus 5.5) with 140 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 41.0% for GPT-4o Mini and 93.2% vs. 40.2% on GPQA Diamond.
Claude Opus 5.5 vs GPT-6.1 Sol
GPT-6.1 Sol is 50.0% cheaper for input tokens ($2.00 vs. $4.00 per 1M tokens) and $10.00 vs. $20.00 for output tokens (2.0x cost advantage). In terms of operational performance, GPT-6.1 Sol delivers lower response latency at 520ms TTFT (160ms faster than Claude Opus 5.5) with 135 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 86.4% for GPT-6.1 Sol and 93.2% vs. 92.4% on GPQA Diamond.
Claude Opus 5.5 vs Grok 3
Grok 3 is 25.0% cheaper for input tokens ($3.00 vs. $4.00 per 1M tokens) and $15.00 vs. $20.00 for output tokens (1.3x cost advantage). In terms of operational performance, Grok 3 delivers lower response latency at 620ms TTFT (60ms faster than Claude Opus 5.5) with 88 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 65.8% for Grok 3 and 93.2% vs. 84.6% on GPQA Diamond.
Claude Opus 5.5 vs Grok 3 Mini
Grok 3 Mini is 92.5% cheaper for input tokens ($0.30 vs. $4.00 per 1M tokens) and $1.20 vs. $20.00 for output tokens (13.3x cost advantage). In terms of operational performance, Grok 3 Mini delivers lower response latency at 150ms TTFT (530ms faster than Claude Opus 5.5) with 160 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 54.0% for Grok 3 Mini and 93.2% vs. 68.0% on GPQA Diamond.
Claude Opus 5.5 vs Llama 3.3 70B Instruct
Llama 3.3 70B Instruct is 97.0% cheaper for input tokens ($0.12 vs. $4.00 per 1M tokens) and $0.30 vs. $20.00 for output tokens (33.3x cost advantage). In terms of operational performance, Llama 3.3 70B Instruct delivers lower response latency at 145ms TTFT (535ms faster than Claude Opus 5.5) with 165 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 42.5% for Llama 3.3 70B Instruct and 93.2% vs. 55.6% on GPQA Diamond.
Claude Opus 5.5 vs Mistral Large 2 (2411)
Mistral Large 2 (2411) is 50.0% cheaper for input tokens ($2.00 vs. $4.00 per 1M tokens) and $6.00 vs. $20.00 for output tokens (2.0x cost advantage). In terms of operational performance, Mistral Large 2 (2411) delivers lower response latency at 380ms TTFT (300ms faster than Claude Opus 5.5) with 75 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 45.2% for Mistral Large 2 (2411) and 93.2% vs. 54.2% on GPQA Diamond.
Claude Opus 5.5 vs o1
Claude Opus 5.5 is 73.3% cheaper for input tokens ($4.00 vs. $15.00 per 1M tokens) and $20.00 vs. $60.00 for output tokens (3.8x cost advantage). In terms of operational performance, Claude Opus 5.5 delivers lower response latency at 680ms TTFT (1920ms faster than o1) with 95 vs. 48 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 48.9% for o1 and 93.2% vs. 78.0% on GPQA Diamond.
Claude Opus 5.5 vs Qwen 3.8 27B
Qwen 3.8 27B is 95.0% cheaper for input tokens ($0.20 vs. $4.00 per 1M tokens) and $0.60 vs. $20.00 for output tokens (20.0x cost advantage). In terms of operational performance, Qwen 3.8 27B delivers lower response latency at 120ms TTFT (560ms faster than Claude Opus 5.5) with 140 vs. 95 tok/s throughput. Claude Opus 5.5 leads coding evaluations at 89.9% SWE-bench Verified vs. 58.4% for Qwen 3.8 27B and 93.2% vs. 68.0% on GPQA Diamond.
Frequently Asked Questions & Query Fan-Out
How much does Claude Opus 5.5 cost per 1M tokens including prompt caching?
Claude Opus 5.5 costs $4.00 per million input tokens and $20.00 per million output tokens. Cached input reads cost $0.200 per 1M tokens, reducing repeated context costs by 95%.
What are Claude Opus 5.5's coding and reasoning benchmark scores?
Claude Opus 5.5 achieves 89.9% on SWE-bench Verified, 97.6% on HumanEval coding, 93.2% on GPQA Diamond graduate-level science, 92.1% on MMLU-Pro, and 98% on MATH-500, earning an overall Inference Value Ratio (IVR) of 57.8/100.
What is the context window, TTFT latency, and token throughput of Claude Opus 5.5?
Claude Opus 5.5 supports a 1,024,000 token context window (128,000 max output tokens) and operates at 680ms average Time to First Token (TTFT) with 95 tokens/sec output streaming throughput.
What are the primary production use cases for Claude Opus 5.5?
Top-ranked frontier intelligence leader for autonomous multi-step agentic coding, repository-scale migrations, and complex scientific reasoning at a 40% lower workload cost than Opus 5.