Skip to content

v0.7.0 — Real-Time Live Benchmark Calculator, Provider Expansion & Discovery Telemetry

Choose a tag to compare

@neuralnexustech neuralnexustech released this 16 Sep 13:07
· 69 commits to main since this release

What's New in v0.7.0

  • ⚡ Live Benchmark Calculator: Added a real-time TTFT and Tokens/sec streaming benchmark calculator directly on provider detail pages (POST /api/providers/[slug]/bench). Measures actual server-side TTFT (Date.now() - t0 upon first content delta) to eliminate browser network jitter, calculates live token throughput, and provides a multi-phase animated execution tracker (connecting → waiting → streaming → done).
  • 🔄 Animated Pulling Models Modal: Replaced silent discovery with an interactive modal and loading animation during upstream catalog pull requests, with an increased 30s timeout and standard browser headers.
  • 🏷️ Model Parameter Scale & Benchmark Scores: Models table and diagnostic cards now show exact parameter sizing (e.g., 32M, 700B) alongside capability benchmark ratings.
  • 🎨 Visual Modality Badges: Upgraded discovery and test cards to display intuitive iconography for text, vision, audio, code, and embedding modalities instead of plain text strings.
  • 💰 Exact Commercial & Free Pricing Normalization: Resolved pricing ambiguity with exact model identifier matching, explicit per-1M token breakdowns, distinct Free tier badges, and human-scale translation.