← All benchmarks
RCM-Bench-1: Claim submission
Health & Medicine · 2026-10-02
Pass rate over three attempts on each of the 20 RCM-Bench-1 scenarios for investigating, correcting and submitting claims, and handling approval and escalation.
Benchmark source: Commure
Top models (higher is better)
| Model | Score | Source |
|---|
| Sonnet 5 | 93.3 | RCM-Bench-1 results: Claude Sonnet 5, Claim submission · Commure |
| Claude Opus 5 | 91.7 | RCM-Bench-1 results: Claude Opus 5, Claim submission · Commure |
| Sonnet 4.6 | 90.0 | RCM-Bench-1 results: Claude Sonnet 4.6, Claim submission · Commure |
| GLM-5.3 | 88.3 | RCM-Bench-1 results: GLM-5.3, Claim submission · Commure |
| Claude Haiku 5.5 | 86.7 | RCM-Bench-1 results: Claude Haiku 5.5, Claim submission · Commure |
| Claude Sonnet 5.5 | 86.7 | RCM-Bench-1 results: Claude Sonnet 5.5, Claim submission · Commure |
| Claude Opus 5.5 | 85.0 | RCM-Bench-1 results: Claude Opus 5.5, Claim submission · Commure |
| Grok 4.7 | 85.0 | RCM-Bench-1 results: Grok 4.7, Claim submission · Commure |
| DeepSeek V4.1 Flash | 81.7 | RCM-Bench-1 results: DeepSeek V4.1 Flash, Claim submission · Commure |
| Gemini 3.8 Flash | 81.7 | RCM-Bench-1 results: Gemini 3.8 Flash, Claim submission · Commure |
| Gemini 4 Argon | 81.7 | RCM-Bench-1 results: Gemini 4 Argon, Claim submission · Commure |
| GPT-5.6 Sol | 81.7 | RCM-Bench-1 results: GPT-5.6 Sol, Claim submission · Commure |
| GPT-5.6 Terra | 80.0 | RCM-Bench-1 results: GPT-5.6 Terra, Claim submission · Commure |
| Opus 4.6 | 78.3 | RCM-Bench-1 results: Claude Opus 4.6, Claim submission · Commure |
| GPT-6 Luna | 78.3 | RCM-Bench-1 results: GPT-6 Luna, Claim submission · Commure |
| GPT-6.1 Sol | 75.0 | RCM-Bench-1 results: GPT-6.1 Sol, Claim submission · Commure |
| Haiku 4.5 | 73.3 | RCM-Bench-1 results: Claude Haiku 4.5, Claim submission · Commure |
| GPT-5 Mini | 73.3 | RCM-Bench-1 results: GPT-5 mini, Claim submission · Commure |
| GPT-6 Sol | 70.0 | RCM-Bench-1 results: GPT-6 Sol, Claim submission · Commure |
| GPT-5.6 Luna | 65.0 | RCM-Bench-1 results: GPT-5.6 Luna, Claim submission · Commure |