GCF is better than JSON for AI agents. Full benchmark data proving it.
JSON was designed in 2001 for browser-to-server communication. When an LLM receives 500 records as JSON, it gets 53,341 tokens of repeated field names ("qualified_name":, "kind":, "score": on every record). At this scale, GPT-5.5 returns empty strings. Claude Opus spends 143 lines manually enumerating symbols and still gets the wrong answer. JSON averages 54.1% comprehension accuracy across 10 models.
GCF declares field names once in a header. Rows are positional values. Section headers (## targets, ## related) replace per-record metadata. The same 500-record payload uses 11,090 tokens instead of 53,341.
- 79% fewer input tokens than JSON
- 63% fewer output tokens than JSON
- 91.2% comprehension accuracy across 10 models and 3 providers (Anthropic, OpenAI, Google)
- Four models hit 100%: Claude Sonnet, Gemini 2.5 Pro, Gemini 3.1 Pro, Gemini 3.5 Flash
- 5/5 generation validity on every frontier model with zero prior training
- Wins all 6 datasets on TOON's own benchmark using their datasets and tokenizer
1,300+ LLM evaluations. 23 comprehension runs. 28 generation runs. 10 models. 3 providers. Deterministic ground truth, no LLM judge, reproducible from one command.
| Model | GCF | TOON | JSON |
|---|---|---|---|
| Claude Opus 4.6 | 96.2% | 84.6% | 73.1% |
| Claude Sonnet 4.6 | 100% | 73.1% | 53.8% |
| Claude Haiku 4.5 | 96.2% | 69.2% | 57.7% |
| GPT-5.5 | 84.1% | 67.7% | 45.8% |
| GPT-5.4 | 78.0% | 56.0% | 44.1% |
| GPT-5.4-mini | 71.8% | 64.1% | 54.2% |
| Gemini 2.5 Pro | 100% | 76.9% | 58.3% |
| Gemini 3.1 Pro | 100% | 76.9% | 46.2% |
| Gemini 3.5 Flash | 100% | 61.5% | 46.2% |
| Gemini 2.5 Flash | 80.6% | 54.1% | 57.0% |
GCF > TOON > JSON on every model from every provider. No exceptions.
| Model | GCF | TOON | JSON |
|---|---|---|---|
| Claude Opus 4.6 | 5/5 | 0/5 | 5/5 |
| Claude Sonnet 4.6 | 5/5 | 2-3/5 | 5/5 |
| GPT-5.5 | 4-5/5 | 1-2/5 | 5/5 |
| GPT-5.4 | 5/5 | 0/5 | 5/5 |
| Gemini 2.5 Pro | 5/5 | 1/5 | 5/5 |
| Gemini 3.1 Pro | 5/5 | 0/5 | 5/5 |
TOON's official decoder rejects LLM-generated output on 7 of 9 models. GCF achieves 5/5 on every frontier model with zero training.
| Format | Tokens (500 symbols) | vs JSON |
|---|---|---|
| GCF | 11,090 | 79% fewer |
| TOON | 16,378 | 69% fewer |
| JSON | 53,341 | baseline |
pip install gcf-proxyWrap any MCP server. Zero code changes. Your server keeps outputting JSON, the model receives GCF.
- Live site
- Full benchmarks
- Full eval data
- GCF Specification
- GCF Proxy
- Playground (live three-way comparison)
- Whitepaper (DOI: 10.5281/zenodo.20579817)
- Implementations: Go | TypeScript | Python | Rust | Swift | Kotlin
MIT - Dayna Blackwell