compare_models
ChatGPTCompare 2–6 specified models/providers using published whole-response speed in one common window. Does not rank answer quality, intelligence or pricing. Provider ambiguity is returned explicitly; no metric substitution.
Overview
Analyze public community measurements of coding-workflow performance. Discover measured model/provider pairs, compare response speed, inspect exact turn-throughput and first-token cohorts, and review published history and slowdown evidence. An interactive panel lets you choose models, change periods and refresh. Results show freshness, sample coverage and measurement limitations. A separate token-use indicator describes observed tokens per request. Public queries require no Tokrate account or desktop installation. Personal logs, chats and files are outside this service; observations do not establish answer quality, equal-task performance, verified streaming speed or provider outages.
Tools
get_efficiency
ChatGPTInspect the published model+reasoning-effort token-efficiency indicator and its reference. It combines tools/providers/versions, and measures observed tokens per request, not answer quality, cost or equal-task benchmarks.
get_history
ChatGPTRead published history for exactly one model (optional provider) or an exact cohortId from get_model. Keeps gaps null and providers/cohorts separate; supports 15m, 1h, 24h, 7d, 30d, never longer retention.
get_methodology
ChatGPTExplain Tokrate response speed, turn throughput, first-token timing, efficiency, publication floors, early-data mode, freshness and detector evidence. Streaming generation speed, quality, intelligence and provider health are not verified.
get_model
ChatGPTInspect one exact model's published response speed and separate tool/provider/effort/version cohorts, recent vs previous periods and history. Optional codingTool/effort filters select cohorts only. Missing observations are unavailable, not zero.
get_signals
ChatGPTInspect current exact-cohort whole-turn-throughput and Codex-reported first-token slowdown evidence and detector readiness. Includes insufficient and stale states. Not official provider incidents; no response-speed detector or proactive notification subscription.
list_models
ChatGPTDiscover exact model/provider identities with published Tokrate community observations. Response-speed rows mix listed coding tools and efforts; cohort-only models have no response-speed value. No personal measurements or quality ranking.
show_comparison
ChatGPTOpen the interactive Tokrate comparison panel. Optionally select 1–6 exact models/providers; otherwise show the first six published response-speed rows. The panel can change its window/selection and refresh. Data is public community observations, not personal logs.
Capabilities
Interactive
Example Prompts
Click any prompt to copy it.
