SambaCloud Release Notes

Release notes for SambaCloud, including new features, enhancements, and fixes.

---

​

October 2025

Release Date: October 7, 2025

​

New Features and Enhancements

​

API Reference Documentation Upgrade

Upgraded the API Reference documentation with a fully automated, Swagger-style interface that dynamically syncs with the OpenAPI specification.

• Interactive, auto-generated API documentation with full model definitions.
• Real-time updates to stay aligned with the latest OpenAPI spec.
• Up-to-date usage examples in Python and TypeScript, powered by the official SambaNova SDKs.

---

​

June 2025

Release Date: June 17, 2025

​

New Features and Enhancements

​

Japanese Language Support

Added Japanese language support for SambaNova documentation.

• Go to the SambaNova documentation homepage.
• Open the language dropdown menu and select Japanese.

---

​

May 2025

Release Date: May 28, 2025

​

New Features and Enhancements

​

AWS Marketplace Integration

SambaCloud is now available as a SaaS offering on AWS Marketplace.

• Subscribe using your AWS account and streamline billing through AWS.
• PrivateLink support for secure, low-latency connections between your AWS VPC and SambaCloud—no public internet exposure.
• Fast onboarding with models like Llama 4, DeepSeek-R1 671B, and Whisper.
• Up to 10x faster inference vs. GPUs with SambaNova’s RDU architecture.
• Privacy-first architecture: SambaNova never stores your data.
• Support for fine-tuned model deployment without code changes.

See AWS Marketplace integration document for more information.

​

DeepSeek-V3-0324 Function Calling Support

Release Date: May 6, 2025 Added function calling support for DeepSeek-V3-0324, enabling more dynamic and programmable interactions by allowing you to invoke external functions directly through the model’s outputs. See Function calling for more information.

---

​

April 2025

​

New Features and Enhancements

​

New Models

• Qwen3-32B (April 29, 2025)
  • Added as a Preview model. High-capacity, multilingual LLM with strong performance across question answering, summarization, reasoning, and coding.
  • Available via Playground and API.
• Whisper-Large-V3 (April 18, 2025)
  • OpenAI’s latest large-scale automatic speech recognition (ASR) model with enhanced transcription accuracy, improved multilingual support, and better robustness against noisy audio.
  • Available via API.
• Llama-4-Maverick-17B-128E-Instruct (April 9, 2025)
  • 400B parameter mixture-of-experts model with 17B active parameters and 128 experts.
  • Added as a Preview model. Competitive with Gemma 3, Gemini 2.0 Flash, and Mistral 3.1.
  • Available via Playground and API.
• Llama-4-Scout-17B-16E-Instruct (April 7, 2025)
  • 109B parameter mixture-of-experts model with 17B active parameters and 16 experts.
  • Added as a Preview model. Competitive with Gemma 3, Gemini 2.0 Flash-Lite, and Mistral 3.1.
  • Available via Playground and API.

​

Llama-4-Maverick Image Input Support

Release Date: April 16, 2025 Llama-4-Maverick-17B-128E-Instruct now supports image input functionality with up to two images as context alongside text prompts. Available via Playground and API.

​

Deprecations

Release Date: April 14, 2025 The following models have been deprecated and are scheduled for removal from active endpoints:

• Llama-3.1-Swallow-70B-Instruct-v0.3
• Llama-3.1-Tulu-3-405B
• Llama-3.2-11B-Vision-Instruct
• Llama-3.2-90B-Vision-Instruct
• Meta-Llama-3.1-70B-Instruct
• Qwen2.5-72B-Instruct
• Qwen2.5-Coder-32B-Instruct

See Model Deprecations for guidance on alternatives.

---

​

March 2025

​

New Features and Enhancements

​

New Models

• DeepSeek-V3-0324 (March 27, 2025)
  • First open-source non-reasoning model that outperforms proprietary non-reasoning models.
  • Major boost in reasoning performance, stronger front-end development skills, and smarter tool-use capabilities.
  • Added as a Preview model. Available via Playground and API.
• E5-Mistral-7B-Instruct (March 20, 2025)
  • Embedding model with Mistral architecture backbone.
  • Recommended for English use. Available via API.
  • See Embeddings capabilities and endpoint documentation.
• QwQ-32B (March 6, 2025)
  • State-of-the-art reasoning model from Alibaba Qwen team.
  • Delivers comparable performance to DeepSeek-R1’s 671B with far fewer parameters.
  • Integrates advanced agent-related features for critical thinking and external tool usage.
• Llama-3.1-Swallow-8B-Instruct-v0.3 and Llama-3.1-Swallow-70B-Instruct-v0.3 (March 5, 2025)
  • Japanese-optimized language models developed through continual pre-training of Meta’s Llama 3.1.
  • Trained on 200B tokens from diverse sources including web corpora, technical content, and multilingual Wikipedia.

​

DeepSeek-R1 Production Release

Release Date: March 18, 2025 DeepSeek-R1 transitioned from preview to production model with increased context length to 16k tokens. Available via Playground and API.

​

API Improvements and Fixes

​

Model List Endpoint

Release Date: March 18, 2025 Added the model list endpoint that provides information about currently available models in SambaCloud.

​

Context Length Increases

• DeepSeek-R1-Distill-Llama-70B (March 13, 2025): Increased to 128k tokens.

---

​

February 2025

​

New Features and Enhancements

​

New Models

• DeepSeek-R1 (February 13, 2025)
  • 671B parameter MoE model with performance comparable to OpenAI’s o1 across mathematics, coding, and reasoning.
  • Added as a Preview model. SambaCloud delivers the fastest DeepSeek-R1 deployment in the world.
• Llama-3.1-Tulu-3-405B (February 4, 2025)
  • Open-source model from Allen Institute for AI (AI2) that performs better than DeepSeek-V3.
  • Trained using Reinforcement Learning with Verifiable Rewards (RLVR).
  • Competitive with or superior to GPT-4o and DeepSeek-V3, with notable advantage in safety benchmarks.
• DeepSeek-R1-Distill-Llama-70B (January 30, 2025)
  • Fine-tuned on Llama 3.3 70B using samples generated by DeepSeek-R1.
  • Outperforms GPT-4o, o1-mini, and Claude-3.5-Sonnet across AIME, MATH-500, GPQA, and LiveCodeBench.

​

Context Length Increases

• Llama-3.3-70B (February 25, 2025): Increased to 128k tokens. Available as production model.
• DeepSeek-R1 (February 21, 2025): Increased to 8k tokens. Available as preview model.

---

​

December 2024

​

New Features and Enhancements

​

New Models

• Llama 3.3 70B (December 11, 2024)
  • Delivers comparable performance to Llama 3.1 405B.
  • Competes closely with OpenAI’s GPT-4o and Google’s Gemini Pro 1.5.
• QwQ 32B Preview (December 11, 2024)
  • Experimental reasoning model from Alibaba’s Qwen team with 32.5B parameters.
  • Excels in mathematics and programming: 65.2% on GPQA, 50.0% on AIME, 90.6% on MATH-500, 50.0% on LiveCodeBench.
• Qwen2.5 72B (December 5, 2024)
  • 72B-parameter model trained on 18T tokens. Supports context lengths up to 128k tokens.
  • Supports 29+ languages including English, Chinese, French, and Spanish.
• Qwen2.5 Coder 32B (December 5, 2024)
  • 32B-parameter model tailored for code-related tasks across 92 programming languages.
  • HumanEval score of 92.7%, matching GPT-4o coding capability.
• Llama Guard 3 8B (December 5, 2024)
  • Fine-tuned for content safety classification, aligning with 14 MLCommons standardized hazards taxonomy.

​

Context Length Increases

• Llama 3.2 1B: Increased from 4k to 16k.
• Llama 3.1 70B: Increased from 64k to 128k.
• Llama 3.1 405B: Increased from 8k to 16k.

---

​

October 2024

​

New Features and Enhancements

​

New Models

• Llama 3.2 11B and 90B (October 29, 2024)
  • Multimodality support for text and image inputs.
• Llama 3.2 1B and 3B (October 1, 2024)
  • Available to all tiers at the fastest inference speed.

​

Function Calling

Release Date: October 29, 2024 Added function calling API enabling dynamic, agentic workflows by allowing the model to suggest and select function calls based on user input.

​

Multimodality in API and Playground

Release Date: October 29, 2024 Interact with multimodal models directly through the Inference API (OpenAI compatible) and Playground for seamless text and image processing.

​

Python and Gradio Code Samples

Release Date: October 29, 2024 New code samples for faster prototyping and reduced setup time.

​

API Improvements and Fixes

​

Automatic Sequence Length Routing

Release Date: October 10, 2024 Automatic routing based on sequence length—no need to change model names to specify different sequence lengths.

​

Context Length Increases

• Llama 3.1 8B (October 10, 2024): Increased from 8k to 16k.
• Llama 3.1 70B (October 10, 2024): Increased from 8k to 64k.

​

Performance Improvements

Improved performance for Llama 3.2 1B and 3B models.

​

User Experience Improvements

• How to Use API guide with example curl code for text and image inputs.
• Streamlined access to updated code snippets.
• New Clear Chat option in Playground.
• New UI components with tool tips.

​

Updated AI Starter Kits

• Multimodal Retriever: Chart, image, and figure understanding with advanced retrieval combining visual and textual data.
• Llama-3.1-Instruct-o1: Enhanced reasoning with Llama-3.1-405B, hosted on Hugging Face Spaces.

---

​

September 2024

Release Date: September 10, 2024

​

New Features and Enhancements

​

SambaCloud Public Launch

• Public launch of the SambaCloud portal, API and the community.
• Access to Llama 3.1 8B, 70B, and 405B at full precision and 10x faster inference compared to GPUs.
• Launched with two tiers: free and enterprise (paid).