๐ฏ Fable vs Fusion Benchmarks
Anthropic Claude Fable 5 ($10/M prompt) vs our Fusion LLMs (as low as $0.15/M)
| Benchmark | Category | Fable 5 | Fusion | Measured | ฮ vs Fable | Comparison |
|---|
๐งช Cost-Constrained Benchmark Testing
Set your budget and we calculate how many real benchmark tests we can run
Budget:
$
Cost: $0.0000
๐ผ Preview
โก Hermes Agent Integration
Add your winning Fusion LLMs as selectable models in Hermes (Dashboard + Telegram)
Once you've benchmarked and validated a Fusion LLM, add it as a custom provider in Hermes. The fusion engine runs as a local API that Hermes calls โ it orchestrates multiple OpenRouter models behind a single endpoint. This means your agent gains the combined intelligence of 3-4 top models at a fraction of Anthropic Fable's cost.
Option 1: Custom Provider
Add fusion as a custom OpenAI-compatible endpoint
hermes config set custom_providers.fusion.base_url http://127.0.0.1:9120/v1
Option 2: OpenRouter Passthrough
Configure as OpenRouter with a fusion model name
hermes model
โ OpenRouter
โ Model: fusion/fusion-phoenix
โ OpenRouter
โ Model: fusion/fusion-phoenix
Option 3: CLI Quick Switch
Test fusion directly from Hermes CLI
hermes chat -q "Build a REST API" \
--model anthropic/claude-fable-5 \
# Then compare with fusion
--model anthropic/claude-fable-5 \
# Then compare with fusion
๐ก Fusion Engine Status
Checking...
๐ฏ Real Benchmark Evaluation (lm-eval-harness)
Run actual lm-evaluation-harness benchmarks against Fusion recipes. Jobs survive disconnects โ check from any device.
Limit:
๐ Development Log
What changed and why โ Fusion LLM changelog