Images processed by multimodal models, rather than images generated.
Image-input counts will appear once providers return normalized multimodal usage consistently.
Observed input-token length across gateway requests over the last 30 days.
Input tokens per request · 1,483 measured requests
Independent model intelligence scores from Artificial Analysis.
20 matched Phaseo models · Higher is better
Claude Opus 5Claude Fable 5
GPT 5.6 Sol
Kimi K3
Qwen 3.8 MaxClaude Opus 4.8
Muse Spark 1.2GPT 5.6 TerraGPT 5.5
Grok 4.5Claude Sonnet 5Claude Opus 4.7Muse Spark 1.1GPT 5.4
GLM 5.2GPT 5.6 Luna
Gemini 3.5 Flash
DeepSeek V4 FlashGemini 3.6 FlashClaude Sonnet 4.6Where Phaseo gateway traffic originated over the last 30 days, based on the country recorded at the gateway edge.
Public applications ranked by model-token usage through Phaseo.
Based on real usage data from Phaseo gateway traffic.
Recent production performance across routes with sufficient public samples.
Highest median output throughput over the last 30 days.
Laguna XS 2.1 (Free)
InklingKimi K3GPT 5.4 NanoGLM 5.2
Macaron V1 Venti (Free)
Leanstral 1.5 (Free)GPT 5.6 LunaGPT Image 2GPT 5.6 TerraClaude Sonnet 4.6Laguna S 2.1 (Free)Gemma 4 31B (Free)Gemma 4 26B A4B (Free)Lowest median response latency over the last 30 days.
Leanstral 1.5 (Free)Laguna XS 2.1 (Free)GPT 5.4 NanoLaguna S 2.1 (Free)GPT 5.6 LunaInklingGemini 3.5 Flash-LiteGemma 4 31B (Free)Gemini 3.5 Flash-LiteGemini 3.5 Flash-LiteClaude Sonnet 4.6Gemini 3.6 Flash
Hy3 (Free)Gemma 4 26B A4B (Free)Kimi K3Text Embedding 3 SmallGPT 5.6 TerraGLM 5.2Macaron V1 Venti (Free)Models ranked by privacy-safe aggregate tool-call volume.
Compare models by tool calls across the selected usage period.
Model adoption by distinct gateway actors, so token-heavy workloads do not dominate the ranking alone.
Compare models by distinct people or workspaces using them across the selected usage period.
