Gemini 3.8 Flash reshapes Agent Arena Pareto frontier at $0.22 per task, 44-50% cheaper
arena · x · 2026-09-03
LMArena reports Gemini 3.8 Flash (High) has landed on the Agent Arena Pareto frontier for the first time, reshaping the cost-performance landscape.
Key numbers:
- Priced at $0.75/$3.75 per MToken (input/output); median cost of $0.22 per task with +5.94% net improvement
- On par with Grok 4.5 (+6.17% at $0.39/task) and GLM 5.2 Max (+6.23% at $0.44/task), but 44-50% cheaper
- Debuted at #14 in Agent Arena, just above DeepSeek-V4-Pro (#15), a big jump from Gemini 3.7 Flash (#32, +0.84%)
- Strong user signals: +14.78% in Praise vs. Complaint, +11.12% in Confirmed Success
- Text Arena: #7 with 1494 pts, ahead of Claude Opus 5 High (#8) and Gemini 3.7 Flash (#10)
More from Models
- Fable 5.1 burns 102K reasoning tokens and hits the 128K output ceiling mid-code — rohanpaul_ai · 2026-09-03
- Google ships five models in one week: Gemini 3.8 Flash, Muse Spark 1.3, and more incoming — sethlazar · 2026-09-03
- Every's writing bench adds Gemini 3.8 Flash, Grok 4.6, and Muse Spark 1.3 — danshipper · 2026-09-03
- Meta's Muse Spark 1.3 lands on OpenRouter with 1M context for agentic workflows — armand_ruiz · 2026-09-03
- DeepSeek-V4-Pro ships with 1.6T-param MoE; open-source eval harness steals the show — DeepLearningAI · 2026-09-03
- Rival AI agents: cross-vendor model review catches what self-review misses — rseroter · 2026-09-03