Best LLMs for JSON / Tool use (2026)
As of 2026-11-14, LLM Reference lists models with function calling or tool use, ordered by release date, newest first, then by name. This list is not a ranking: Tell-1 by Retell AI is listed first because it has the newest release date on the list, 2026-11-13.
Last refreshed 2026-11-14. Next refresh: weekly.
LLMs for JSON / Tool use in 2026, shown unranked and sorted by release date, newest first. Includes native JSON output and structured output support.
How we rank
This page is an unranked list of models listed with function calling or tool use, sorted by release date, newest first. The BFCL board has not updated since April 2026, so no score sets the order.
- Eligibility — Models listed with function calling or tool use.
- Display order — An unranked list with no podium. Models are sorted by newest release, then by name. BFCL (V4) and τ-bench scores are shown for reference where they exist; they do not set the order.
- Variant collapse — We keep one row per model family (`familySlug` + parameter tier). This list is unranked, so the row we keep is the sibling that comes first in the list's own display order (see Display order above); if siblings tie on that order, the newest `release` is kept, then the model name and slug decide. Benchmark scores shown beside a model do not choose which sibling is kept.
- Pricing — Lowest tracked input(output) for API routes.
JSON / Tool use models
Unranked. Newest release first. The BFCL board has not updated since April 2026.
| Model | Input $/1M | Output $/1M | |
|---|---|---|---|
| Tell-1 PreviewReasoningTools BFCL: — | — | — | |
| Gemini 4 Hydrogen ReasoningVisionTools BFCL: — | $1.50 | $7.50 | |
| ERNIE 5.5 Flash Preview PreviewReasoningTools BFCL: — | Free | Free | |
| Qwen4-Max-Preview PreviewReasoningVisionTools BFCL: — | — | — | |
| Raon-1 nano ReasoningTools BFCL: — | — | — | |
| Nex-N2.6 ReasoningVisionTools BFCL: — | — | — | |
| Claude Fable 5.2 ReasoningVisionTools BFCL: — | $8.00 | $40.00 | |
| Claude Mythos 5.2 RestrictedReasoningVisionTools BFCL: — | — | — | |
| K-EXAONE 2.5 ReasoningTools BFCL: — | — | — | |
| Muse Ember ReasoningVision BFCL: — | $0.10 | $0.50 | |
| GPT-6 Cyber ReasoningTools BFCL: — | — | — | |
| Inkling 2 Small ReasoningVisionTools BFCL: — | Free | Free | |
| Ling-3.2-flash VisionTools BFCL: — | — | — | |
| Ripple-65B Preview PreviewTools BFCL: — | $0.20 | $0.60 | |
| SWE-2 Flash ReasoningTools BFCL: — | — | — | |
| LongCat-2.5 ReasoningVisionTools BFCL: — | — | — | |
| Phonic-1 Tools BFCL: — | — | — | |
| Gemini 4 Carbon ReasoningVisionTools BFCL: — | $2.00 | $10.00 | |
| Particle-1 ReasoningVisionTools BFCL: — | $0.25 | $1.00 | |
| LFM3-1.7B Tools BFCL: — | $0.08 | $0.20 |
Browse Other Categories
Frequently asked questions
Which LLMs support function calling and tool use?
As of 2026-11-14, the newest models on this list of models with function calling or tool use are Tell-1 by Retell AI (released 2026-11-13), Gemini 4 Hydrogen by Google DeepMind (released 2026-11-12) and ERNIE 5.5 Flash Preview by Baidu AI (released 2026-11-11). The list is sorted by release date, so the order is not a quality ranking.
How is this list ordered?
By release date, newest first, then by model name. BFCL and τ-bench scores do not affect the order.
Which models are included?
Models listed with function calling or tool use that are not deprecated and have at least one active provider route. One model is kept per family and size tier.
What do the BFCL and τ-bench columns show?
Benchmark scores shown for reference only; neither column sets the order. BFCL is the Berkeley Function-Calling Leaderboard, and its column shows V4 scores only. Scores from earlier BFCL versions and self-reported BFCL scores are not shown. A dash means no score is listed for that model.
How often is this page updated?
Weekly. Last refreshed 2026-11-14.