What exactly changed in AI models: context, prices, capabilities. Only values stated by sources — never guessed.
| Parameter | Before | Now |
|---|---|---|
| WER (Najdi+Hijazi test) | 55.05% | 29.96% |
| CER (Najdi+Hijazi test) | 31.63% | 12.18% |
| Full SADA WER | 58.84% | 35.61% |
| FLEURS English WER | 11.04% | 10.42% |
| FLEURS Arabic WER | 12.67% | 11.41% |
| Trainable parameters (top-8 layers unfrozen) | Not specified | 230.4M of 638M total |
| Language coverage | 40 language-locales (base model) | Same, plus specialized Najdi/Hijazi dialects |
| Parameter | Before | Now |
|---|---|---|
| Base model | Not specified | Qwen3-8B |
| Report generation time (avg, full pipeline) | 178.5 seconds (Claude-powered Thinking mode) | 51.1 seconds (AstaBrief Fast mode) |
| Speed vs Thinking mode | 1x (baseline) | ~3.5x faster |
| SFT training examples | Not specified | 47K (from 90K filtered research queries) |
| DPO training examples | Not specified | ~6K |
| Weights/Data availability | Proprietary (Claude 3.5/3.7 Sonnet, o3, o4-mini, GPT-4.1 used for data generation) | Open-sourced model weights and training data |
| Parameter | Before | Now |
|---|---|---|
| Execution mode | Cloud-only agentic workflows | Local/offline execution via LiteRT with Gemma 4 26B A4B, plus hybrid cloud-local orchestration |
| Hardware requirement | Not specified | Recommended >24GB VRAM or unified memory |
| Cost in hybrid demo | Not specified | 97.2% of tokens (3,322) run locally/offline; only 95 cloud tokens used for Gemini 3.8 Flash planning |
| Compatible backends | Not specified | Supports OpenAI-compatible local servers (Ollama, LM Studio, vLLM) via LocalOpenAIAgentConfig |