AI Frameworks · ShishirPatil

gorilla

Gorilla is an open-source project that fine-tunes large language models to accurately invoke APIs and call functions from natural language instructions. It includes pre-trained models, a comprehensive function-calling leaderboard (BFCL), evaluation datasets, and a runtime engine (GoEx) for safely executing LLM-generated actions.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/ShishirPatil/gorilla

12.9k

GitHub stars

1.4k

Forks

Python

Primary language

Apache-2.0

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	ShishirPatil/gorilla
Owner	ShishirPatil
Primary language	Python
License	Apache-2.0 — OSI-approved
Stars	12.9k
Forks	1.4k
Open issues	264
Latest release	v1.3 (2025-07-17)
Last updated	2026-04-13
Source	https://github.com/ShishirPatil/gorilla

What gorilla is

Gorilla provides fine-tuned LLM checkpoints optimized for function calling across 1,600+ APIs, retrieval-augmented training infrastructure, multi-turn evaluation frameworks, and a post-facto validation runtime (GoEx) that supports OAuth2 authentication and undo/damage-confinement abstractions for autonomous agent execution.

Quickstart

Get the gorilla source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/ShishirPatil/gorilla.git2cd gorilla3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Building Tool-Calling / Function-Calling Assistants

Use Gorilla models or OpenFunctions-V2 to build chatbots and agents that reliably invoke APIs, external services, or internal tools from user queries. Includes support for parallel execution and multiple data types.

Evaluating and Benchmarking Function-Calling Capabilities

Leverage the Berkeley Function Calling Leaderboard (BFCL V3+) to benchmark your own models or fine-tuned variants against standard datasets that cover single-turn, multi-turn, multi-step, and agentic scenarios with cost/latency metrics.

Deploying Autonomous LLM Agents with Safety Guardrails

Use GoEx runtime to execute LLM-generated API calls, code, and actions with post-execution validation, undo capabilities, and damage confinement to mitigate unintended side effects in production agentic systems.

Implementation considerations

Choose between pre-trained Gorilla models (7B/13B), fine-tuned OpenFunctions-V2 for multi-language support, or integrate the BFCL evaluation pipeline into your own training workflow.
API documentation ingestion and retrieval is central to Gorilla's design; plan for curating or indexing your target APIs (1,600+ in APIBench) and handling documentation drift over time.
Fine-tuning recipes and evaluation code are provided but assume familiarity with LLM training (transformer stacks, LoRA, etc.); budget for compute and hyperparameter tuning.
GoEx runtime requires careful configuration of OAuth2 credentials, API key management, and undo/rollback logic; test post-facto validation rules thoroughly before autonomous execution.
Integrate with your existing LLM inference stack (vLLM, TensorFlow Hub, Torch Hub are supported); compatibility with your serving infrastructure should be verified early.

When to avoid it — and what to weigh

Requiring Out-of-the-Box Closed-Source Model Quality — Gorilla models are open-source and smaller (7B–13B variants noted); if you need the raw capability of GPT-4 or Claude, proprietary APIs may be more direct, though Gorilla excels at function calling specifically.
Needing Comprehensive Commercial Support & SLA — Gorilla is community-driven research software. While Apache 2.0 permits commercial use, there is no vendor-backed SLA, bug-fix guarantees, or commercial support contracts mentioned in the repository.
Deploying in Strict Offline or Air-Gapped Environments — Gorilla relies on external API documentation retrieval and leaderboard infrastructure; offline-first or highly restricted network deployments may require significant custom adaptation.
Low Tolerance for Active Issue Backlogs — The repository shows 264 open issues; if you require a bug-free, production-hardened codebase, verify issue severity and resolution timelines match your SLA expectations.

License & commercial use

Apache License 2.0. Permissive OSI license permitting commercial use, modification, and distribution with appropriate attribution and liability disclaimer.

Apache 2.0 explicitly permits commercial use. However, no vendor support, warranties, or liability protections are provided by the project maintainers. Review the full Apache 2.0 license terms and conduct your own security/compliance audit before deploying in production. Consider whether you need commercial support or insurance before relying on this for mission-critical systems.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Strong
Assessment confidence	High

Security considerations

Gorilla and GoEx handle API invocation and credential management; key concerns include: (1) API key/OAuth2 secret storage and rotation—ensure compliance with your credential management policy; (2) Post-facto validation in GoEx assesses actions after execution, so rollback logic must be tested; (3) No mention of formal security audits or vulnerability disclosure process in provided data; (4) Open-source model weights are public; verify no sensitive training data is exposed. Conduct a formal threat model if integrating GoEx for autonomous agent execution.

Alternatives to consider

OpenAI GPT-4 / Claude function-calling APIs

Closed-source, vendor-supported, higher baseline capability. No fine-tuning or local deployment; higher per-token cost. Best if you prioritize reliability and support over customization.

LangChain / LlamaIndex function-calling abstractions

Framework-level tooling for routing LLM calls to external APIs; agnostic to model choice. Lighter weight than Gorilla but less specialized for function-calling optimization and evaluation.

Vellum / Humanloop (managed platforms)

Closed-source, vendor-managed evaluation and deployment of function-calling agents. Premium support and monitoring. Trade-off: higher cost, less transparency, no local fine-tuning.

Software development agency

Build on gorilla with DEV.co software developers

Explore Gorilla's pre-trained models, evaluate against BFCL, or integrate GoEx into your agentic pipeline. Start with Hugging Face checkpoints or run the local evaluation suite.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

marvin END-TO-END-GENERATIVE-AI-PROJECTS generative-ai

Related on DEV.co

Explore the category and the services that help you build with it.

AI Frameworks AI Application Development API Development Enterprise Software Talk to DEV.co

gorilla FAQ

Can I use Gorilla models commercially?

Yes, Apache 2.0 permits commercial use. However, you assume all liability and responsibility for security, compliance, and performance. Verify Gorilla meets your security and regulatory requirements before production deployment.

Do I have to fine-tune Gorilla, or can I use pre-trained weights?

Pre-trained Gorilla and OpenFunctions-V2 models are available on Hugging Face and can be used out-of-the-box via standard inference servers. Fine-tuning is optional and recommended if your API set or domain differs significantly from APIBench.

What is GoEx and why would I use it?

GoEx is a runtime that executes LLM-generated actions (API calls, code) with post-execution validation, undo, and damage confinement. Use it if you want to run autonomous agents with risk mitigation and rollback capabilities.

How do I evaluate my model against BFCL?

The Berkeley Function Calling Leaderboard repository includes evaluation code and datasets (V1–V4). Run your model predictions through their evaluation script to get cost, latency, and accuracy metrics comparable to public leaderboard results.

Work with a software development agency

From first prototype to production, DEV.co delivers software development services around tools like gorilla. Our software development agency staffs experienced software developers and web developers for custom software development, web development, integrations, and ongoing support across ai frameworks and beyond.

Ready to Build Tool-Calling Agents?

Explore Gorilla's pre-trained models, evaluate against BFCL, or integrate GoEx into your agentic pipeline. Start with Hugging Face checkpoints or run the local evaluation suite.

Talk to DEV.co Schedule a Consultation