Workload (input : output ratio)
Latency sensitivity
Your location
AI Value = log₁₀(quality / blended $) − log₁₀((ms + k/4) / (k + k/4))
ms = TTFT + city→region RTT + α·gen_time · bonus below knee k, penalty above · +1.0 ≈ 10× value
Cheapest Input
—
per 1M tokens
Best AI Value (chat)
—
AI value score
Fastest from Vilnius 🇱🇹
—
estimated RTT
Best Reasoning AI Value
—
AI value score
EU-Native Models
—
with European endpoints

🏆 Best AI Value per Category

Loading…
Provider ↕ Model ↕ In / 1M ↕ Out / 1M ↕ Ctx ↕ Latency 🇱🇹 ↕ Quality Index ↕ AI Value ↕ Links
Loading models…
📊 Methodology: Quality = Artificial Analysis Intelligence Index (rebased 2026, top frontier ~57–60). Cost = workload-weighted blend of input & output $/1M tokens. Latency = TTFT (inference) + estimated RTT from your selected city to the model's datacenter region, then a symmetric log adjustment around the sensitivity knee (bonus below, penalty above). Throughput (tok/s) is shown per model and folded into typical response time for the chosen workload.
Formula: AI Value = log₁₀(quality_index / blended_cost) − log₁₀((latency_ms + knee/4) / (knee + knee/4)). Absolute score — no rescaling. LMSYS Chatbot Arena ELO shown separately for human-preference context. Auto-refreshed every 6h.