Coding & Dev 📅 2026-11-18 · Mini AI Tools Research Desk

Match a SambaNova Call to an Id the Live List Still Has

On 18 Nov 2026, GET /v1/models returned four ids. The quickstart still names Meta-Llama-3.3-70B-Instruct, which was not among them.

$ EXECUTIVE BENCHMARK SUMMARY (TL;DR)

Copy an id the live list still returns

On 18 November 2026, GET https://api.sambanova.ai/v1/models returned HTTP 200 and four ids: GLM-5.3, MiniMax-M3.1-Flash, gemma-4-31B-it, and gpt-oss-120b. No API key was sent with that request. The quickstart still uses Meta-Llama-3.3-70B-Instruct in its samples. That id was not in the four. The docs model page still lists DeepSeek-V3.1 and DeepSeek-V3.2. Those two were absent as well.

Context length on the live list was 1,048,576 for GLM-5.3 and MiniMax-M3.1-Flash, and 131,072 for the other two. The docs page calls 131,072 tokens 128k. It also says gemma-4-31B-it takes text, image, and video, and that audio input is not supported. GLM-5.3 does not appear on that docs page. https://docs.sambanova.ai/docs/en/models/model-deprecations returned HTTP 404. That path did not load a deprecation notice, and this page does not treat the missing ids as officially deprecated.

Four steps: read the live list of four ids, drop Llama 3.3 70B, keep gpt-oss-120b and gemma-4-31B-it, note the GLM-5.3 dashboard sample🔍 View full resolution
The live list returned four ids on 18 Nov 2026. The quickstart model name was not one of them.

A script that pastes the quickstart model name calls an id the live list did not return that day.

Live list: api.sambanova.ai/v1/models

Quickstart: docs.sambanova.ai/docs/en/get-started/quickstart

Model page: docs.sambanova.ai/docs/en/models/sambacloud-models

Create the key, then send one chat call

The keys page says each user can generate and use up to 25 API keys, and that a key cannot be viewed again. The portal at /apis rendered the same cap on 18 November 2026: "You can create a maximum of 25 API keys." The base URL printed for an OpenAI client is https://api.sambanova.ai/v1. Chat calls go to https://api.sambanova.ai/v1/chat/completions. The header is Authorization: Bearer plus the key.

The quickstart Python sample sets base_url to the placeholder your-sambanova-base-url and the model to Meta-Llama-3.3-70B-Instruct. Put https://api.sambanova.ai/v1 in the base URL. Put an id from the live four in the model field. The public dashboard HTML samples GLM-5.3 on that chat URL, with stream set to true. The playground page returned HTTP 200, and its HTML named GLM-5.3 too. A body that stays on the live list looks like this: POST https://api.sambanova.ai/v1/chat/completions, header Authorization: Bearer plus the key, JSON field model set to gpt-oss-120b, and messages with one user line. That POST was not sent.

Get a key: cloud.sambanova.ai/apis

Try it: cloud.sambanova.ai/playground

20 a minute on two ids, and a plans card that asks for a card

The rate-limit page says the free tier applies when no payment method is linked, and the developer tier applies when a payment method is linked. Each free-tier row is 20 requests a minute, 20 requests a day, and 200,000 tokens a day. The five ids in those rows are DeepSeek-V3.1, Meta-Llama-3.3-70B-Instruct, gpt-oss-120b, DeepSeek-V3.2, and gemma-4-31B-it. Only gpt-oss-120b and gemma-4-31B-it were also in the live four. GLM-5.3 and MiniMax-M3.1-Flash were on the live list and were not in the free-tier tables, so that 20 / 20 / 200,000 line was not printed for them.

Free tier is 20 requests a minute, 20 a day, and 200,000 tokens a day when no card is linked. The plans card says to add a card and buy credits.🔍 View full resolution
On 18 Nov 2026 the rate-limit page described a no-card free tier, and the plans page told readers to add a card first.

The plans page Free card says to add a payment method and purchase credits to run your first requests. It also lists pay-as-you-go credits and access to production models. Both official pages were fetched the same day. This check did not log in, so it did not see whether the console blocks the first call. The plans sentence is a printed instruction, not a block this page observed.

With a payment method linked, the rate-limit page caps a developer-tier account at 20 million tokens a day across all models. On that tier, gpt-oss-120b and gemma-4-31B-it are listed at 60 requests a minute and 12,000 requests a day. The pricing page labels its columns per 1 million tokens, in USD. Input / output there is 2.80 / 8.80 for GLM-5.3, 0.30 / 1.20 for MiniMax-M3.1-Flash, 0.38 / 1.15 for gemma-4-31B-it, and 0.22 / 0.59 for gpt-oss-120b. Cached input is 0.14 and 0.03 for the first two, and N/A for the last two.

IdOn the live listFree-tier rowInput / output USD per 1M
gpt-oss-120bYes20/min, 20/day, 200,000 tokens0.22 / 0.59
gemma-4-31B-itYes20/min, 20/day, 200,000 tokens0.38 / 1.15
GLM-5.3Yes. Context 1,048,576No row2.80 / 8.80
MiniMax-M3.1-FlashYes. Context 1,048,576No row0.30 / 1.20
Meta-Llama-3.3-70B-InstructNoRow printed, id absent from the fourNot on the pricing page

Rate limits: docs.sambanova.ai/docs/en/models/rate-limits

Plans: cloud.sambanova.ai/plans

Prices: cloud.sambanova.ai/plans/pricing

Which id to paste

Choose gpt-oss-120b or gemma-4-31B-it if the call needs an id that is on the live list and also has a free-tier row: 20 requests a minute, 20 a day, 200,000 tokens a day. Recheck /v1/models the day you send it.

Live list: api.sambanova.ai/v1/models

Choose GLM-5.3 if you are following the dashboard sample. It was on the live list, with context length 1,048,576. The free-tier table did not name it. The pricing page prints 2.80 USD input and 8.80 USD output per 1 million tokens.

Try it: cloud.sambanova.ai/playground

Choose MiniMax-M3.1-Flash if you want the docs preview row that was still on the live list. The docs call the window 1M tokens, text and image. The live field was 1,048,576, which is that same 2^20 count. The free-tier table did not name this id. The pricing page prints 0.30 USD input and 1.20 USD output per 1 million tokens.

Model page: docs.sambanova.ai/docs/en/models/sambacloud-models

Read the plans page if you need the printed line that says to add a payment method and purchase credits. The rate-limit page describes a no-card free tier for the ids in its table. This check did not log in, and it did not watch the console stop a request.

Plans: cloud.sambanova.ai/plans

Get a key: cloud.sambanova.ai/apis

Leave Meta-Llama-3.3-70B-Instruct, DeepSeek-V3.1, and DeepSeek-V3.2 out of the body until GET /v1/models returns them again.

Featured Tools in this Guide

CW

Cloudflare Workers AI

Coding & Dev
Freemium

Create a Workers AI token and send one REST prompt to run Llama 3.1 8B on Cloudflare.

OL

Ollama

Coding & Dev
Freemium

Install Ollama, pull gemma4:e2b, and chat with the model on your own computer.

GR

Groq

Coding & Dev
Freemium

High-speed LLM inference API powered by LPU hardware, offering free developer quotas for open-weight and speech models.

OP

OpenRouter

Coding & Dev
Freemium

Unified AI model router providing single-API access to hundreds of commercial and open-weight models, including free tiers.

More In-Depth Benchmarks & Guides

View all guides →