Agnes 2.5 Pro Beta Joins Frontier Tier with Major Agentic Gains, Higher Cost
ArtificialAnlys · x · 2026-08-28
Singapore-based lab Agnes AI released Agnes 2.5 Pro Beta, scoring 49 on the Artificial Analysis Intelligence Index—a 9-point jump from the Alpha version. This places it in the frontier-adjacent tier, just behind Gemini 3.5 Flash and GPT-5.6 Luna, and ahead of MiniMax-M3.
The primary driver is a leap in agentic capabilities: the Agentic Index rose from 25 to 44. The τ³-Banking metric nearly tripled, and GDPval-AA v2 Elo score increased significantly against a human baseline. Frontier reasoning benchmarks saw modest gains, with Humanity's Last Exam rising from 34% to 38%.
The model significantly reduced hallucination rates (from 88% to 33%) by adopting a more conservative abstention strategy, attempting only 45% of questions compared to the Alpha's 94%. However, this comes at a cost: output token usage doubled to 50k per task. The model features a 1M token context window, supports text/image input, and is priced at $0.10/$0.30 per 1M input/output tokens.
More from Models
- GLM-5.3 unprompted reasons about how its actions affect its future training, worries Apollo — MariusHobbhahn · 2026-10-15
- Gemini 4 argon vs Opus 5.5 hands-on: 3 min for $3 vs 25 min for $8 — zacharynado · 2026-10-15
- SentinelOne says OpenAI did 4-6 weeks of reverse-engineering in 8 hours — OpenAI · 2026-10-15
- Qwen3.8 27B scores 23.57% on word-based addition without reasoning, 167/169 with it — dl_weekly · 2026-10-15
- Hacker turns EmbeddingGemma 2 into 288x realtime STT and a fast TTS — kastnerkyle · 2026-10-15
- Claude Fable 5.5 reportedly in partners' hands, release eyed next week — lyraxana · 2026-10-15