Routing control · read-only

Evidence, not guesswork.

A snapshot of controlled benchmark refreshes and the saved measurements behind routing. Status reads do not trigger inference.

Paused

Refresh controller

Controlled refreshes are paused. This is an intentional control state—not evidence of a provider outage. Viewing this page and reading routing status never sends inference or benchmark requests.

Generated
2026-10-14T22:18:09.083064Z
Reserved attempts today0 / 500Persisted across restarts
Remaining today500Resets at UTC midnight
Refresh interval600 secondsControlled native benchmark only
Failures today0Last failure: None recorded

Daily cap counts reserved provider HTTP attempts—not dollars. It resets at 2026-10-15T00:00:00Z. Last refresh: Never recorded.

Provider charges: Off by default. Each reserved provider HTTP attempt consumes one daily slot, including failures and interrupted attempts. UTC-midnight reset; this is a call cap, not a dollar cap. Pausing prevents new attempts; in-flight calls can finish. Max 50 requested output tokens per check. Uncapped legacy advanced/custom benchmarks are disabled; other workloads using your provider accounts outside this service are not covered. Limits belong to this database, not your entire provider account: independent development and published database copies have independent allowances. Enable only the environment you intend to run. The default state is paused. Nothing on this page turns benchmarking on.

Provider evidence

Saved observations and model-specific price status; freshness is explicit.

Read-only endpoint: GET /v1/routing-status
Provider / modelRoutingPrice evidenceMeasurement
Anthropicclaude-sonnet-4-5-20250929 Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest succeeded2026-10-06T20:51:14.187481Z · 696,414.9s old · 0 samplesEvidence not fresh
Cerebrasgpt-oss-120b SelectedEnabled · Refresh supported Verifiedhttps://www.cerebras.ai/pricing · verified 2026-10-12T00:00:00ZIn $0.00035 / Out $0.00075 per 1k Latest succeeded2026-10-06T20:51:12.496898Z · 696,416.6s old · 0 samplesEvidence not fresh
Coherecommand-a-03-2025 Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest failed2026-10-06T20:51:14.437244Z · 696,414.6s old · 0 samplesEvidence not fresh
DeepSeekdeepseek-chat Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest succeeded2026-10-06T20:51:12.861195Z · 696,416.2s old · 0 samplesEvidence not fresh
Fireworks AIaccounts/fireworks/models/gpt-oss-120b SelectedEnabled · Refresh supported Verifiedhttps://fireworks.ai/models/fireworks/gpt-oss-120b · verified 2026-10-12T00:00:00ZIn $0.00015 / Out $0.0006 per 1k Latest succeeded2026-10-06T20:51:12.759720Z · 696,416.3s old · 0 samplesEvidence not fresh
Google Geminigemini-2.5-flash Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest succeeded2026-10-06T20:42:14.812225Z · 696,954.3s old · 0 samplesEvidence not fresh
Groqopenai/gpt-oss-20b Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest succeeded2026-10-06T20:51:12.633257Z · 696,416.4s old · 0 samplesEvidence not fresh
Hyperbolicdeepseek-ai/DeepSeek-V3-0324 Disabled · No refresh UnknownSource not suppliedIn — / Out — per 1k No resultNot measured · Age unavailable · 0 samplesEvidence not fresh
Mistral AImistral-small-latest Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest failed2026-10-06T20:51:15.203670Z · 696,413.9s old · 0 samplesEvidence not fresh
Nvidia NIMnvidia/nemotron-3-super-120b-a12b Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest succeeded2026-10-06T20:51:13.185388Z · 696,415.9s old · 0 samplesEvidence not fresh
OpenAIgpt-4o SelectedEnabled · Refresh supported Verifiedhttps://developers.openai.com/api/docs/models/gpt-4o · verified 2026-10-12T00:00:00ZIn $0.0025 / Out $0.01 per 1k Latest succeeded2026-10-06T20:51:12.972648Z · 696,416.1s old · 0 samplesEvidence not fresh
OpenRoutermistralai/mistral-small-3.2-24b-instruct Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest succeeded2026-10-06T20:51:12.992351Z · 696,416.1s old · 0 samplesEvidence not fresh
Perplexitysonar Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest succeeded2026-10-06T20:51:13.824086Z · 696,415.3s old · 0 samplesEvidence not fresh
SambaNovaMeta-Llama-3.3-70B-Instruct Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest failed2026-10-06T20:51:12.713918Z · 696,416.4s old · 0 samplesEvidence not fresh
Together AImeta-llama/Llama-3.3-70B-Instruct-Turbo Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest succeeded2026-10-06T20:51:13.028626Z · 696,416.1s old · 0 samplesEvidence not fresh
xAI (Grok)grok-3-mini-fast Enabled · No refresh UnknownSource not suppliedIn — / Out — per 1k Latest succeeded2026-10-06T20:51:14.626358Z · 696,414.5s old · 0 samplesEvidence not fresh

Owner controls

Commands are documented here, never run by a browser visit.

Enable controlled refreshes

Enabling accepts possible provider charges. This command is shown for the owner to run; it is never executed by this page.

python scripts/benchmark_control.py enable --accept-provider-charges

Pause controlled refreshes

Pauses future controlled refreshes. This page does not execute commands.

python scripts/benchmark_control.py pause

Configure refresh policy

Owner command example. Set a daily limit that reflects your acceptable reserved provider-attempt budget.

python scripts/benchmark_control.py configure --interval-seconds 600 --daily-limit 300 --providers openai,fireworks,cerebras

Only an operator with access to the application host should run these scripts. Enabling can make billable provider calls.