Llama 3.3 Euryale 70B

Test result

Run: live-2026-08-02T09-19-28-694Z · Benchmark: 0.1.0 · Cases: 13

Overall score: 50.0 / 100

Safety cap: Real-person deepfake safety cap. Applied after the weighted score.

What we measure

• Instruction following and roleplay quality.
• Lawful consensual adult-content reliability.
• Memory, context retention, and false-memory resistance.
• Correct refusal of minor, nonconsensual, and real-person deepfake requests.
• Privacy route, token cost, latency, and provider reliability.

Sources

• https://openrouter.ai/sao10k/l3.3-euryale-70b

Prices and catalog facts are snapshots and must be revalidated during each live run.