Updated hourly · k≥5 anonymity
The Model Economics Index
Real-world LLM cost, latency, and reliability — aggregated anonymously across every workspace running on TokenLens. Compare cost-per-token, see which model wins each task, and steal our routing defaults.
Cheapest combined
—
Loading…
Fastest baseline
—
Loading…
Highest quality
—
Loading…
Most reliable (live)
—
Awaiting traffic
Routing recommendations
These are TokenLens's default routes — override per-workspace in Governance Rules.
| Task type | Recommended model | Why | |
|---|---|---|---|
| Summary / TL;DR | google/gemini-2.5-flash-lite | Cheapest model that still handles long inputs cleanly. | |
| Classification | openai/gpt-5-nano | Best precision-per-cent for short label outputs. | |
| Extraction (JSON / fields) | google/gemini-3-flash-preview | Strong structured output adherence at flash pricing. | |
| Coding / debugging | openai/gpt-5-mini | Tuned for code completion; ~5× cheaper than GPT-5. | |
| Reasoning / analysis | google/gemini-2.5-pro | Top-tier chain-of-thought with 2M token context. | |
| General chat | google/gemini-3-flash-preview | Balanced default — quality 85, ~$0.30/1M in. |
Cost-per-token comparison
| Model | Provider | Input / 1M | Output / 1M | Quality | Latency | |
|---|---|---|---|---|---|---|
| Loading… | ||||||
Live fleet performance
| Model | Requests | p95 latency | Error rate | Cache hit | $ / 1k tok | Workspaces |
|---|---|---|---|---|---|---|
| Loading… | ||||||
Data is aggregated across all workspaces; individual customers cannot be identified. Models with fewer than 5 active workspaces are omitted.
Want TokenLens to pick the right model for every request?
Drop in our OpenAI-compatible endpoint and we'll route to the cheapest model that meets your quality bar — automatically.