Agent Arena: Gemini 4 Carbon Gets More Cost-Competitive as Tasks Get Harder
rseroter · x · 2026-11-11
Agent Arena's latest agent-cost analysis finds Google's Gemini 4 Carbon grows more cost-competitive as tasks become longer and more demanding.
- On a typical task, GPT-6 Astra costs about 3.4× as much as Carbon, Claude Opus 5.5 about 3.9×, and GPT-6.1 Sol is the cheapest option at roughly 0.54×.
- Across the 5% most expensive sessions, Carbon improves against all three rivals: Opus edges up to 4.2×, Astra's relative cost roughly doubles, and Sol's discount shrinks.
- The analysis notes these gaps translate into meaningful real-world savings on long tasks such as building an Android security app.
More from Infra
- SpaceX locks 100+ GW of solar cell equipment from SINGULUS for Starmind AI satellites — ns123abc · 2026-11-11
- CoreWeave upgrades to CNCF Platinum member, runs its GPU cloud entirely on Kubernetes — altryne · 2026-11-11
- SpaceX teams with Qualcomm to build AI-native base stations, eyeing first AI carrier — elonmusk · 2026-11-11
- Prism capability extraction: specialists carved from Qwen 27B at <20% memory match full model — tokenbender · 2026-11-11
- Apple reportedly building Harbor app to turn Macs into private AI inference servers with Thunderbolt 5 clustering — BenBajarin · 2026-11-11
- AMD Zen 6 deep dive: cheaper branch mispredicts and a return to Zen 3/4 integer scheduling — jedisct1 · 2026-11-11