Skip to content

Compare models

Put up to 4 models beside each other — token prices, context windows, capabilities and provider, from the same catalogue the model pages read.

  1. GPT OSS 120BOpenAIRemove
  2. Gemini 4 CarbonGoogleRemove
  3. Nemotron 3.5 Super VLNVIDIARemove
gpt-oss-120b vs gemini-4-carbon vs nemotron-3.5-super-vl
AttributeGPT OSS 120Bgpt-oss-120bGemini 4 Carbongemini-4-carbonNemotron 3.5 Super VLnemotron-3.5-super-vl
Pricing
Input$0.037 / 1M$2.00 / 1M$0.10 / 1M
Output$0.17 / 1M$10.00 / 1M$0.40 / 1M
Cache Write (5m)$0.037 / 1M$2.00 / 1M$0.10 / 1M
Cache Write (1h)$0.037 / 1M$2.00 / 1M$0.10 / 1M
Cache Read$0.037 / 1M$2.00 / 1M$0.10 / 1M
Web Search$0 / 1M$0 / 1M$0 / 1M
Context
Max context131.1K1M1M
Max outputN/AN/AN/A
Capabilities
VisionNoYesYes
Function CallingYesYesYes
JSON ModeYesYesYes
StreamingYesYesYes
Catalogue
ProviderOpenAIGoogleNVIDIA
Categorychatchatchat
Charge typePay As You GoPay As You GoPay As You Go
Released———
Description
Summarygpt-oss-120b is an open-weight 117B-parameter MoE model from OpenAI, built for advanced reasoning and production workloads. Only about 5.1B parameters are active per step, and it’s optimized to run on a single H100 using MXFP4 quantization. It supports adjustable reasoning depth, full chain-of-thought, and native agent features like tool use, function calling, browsing, and structured outputs.Gemini 4 Carbon is Google's frontier multimodal reasoning model, designed for advanced coding, agentic workflows, and complex multi-step problem solving. It supports text, image, audio, video, and file inputs, enabling integrated reasoning and analysis across diverse modalities. Optimized for long-horizon execution and demanding professional workloads, Gemini 4 Carbon supports up to 1M output tokens through continuation and configurable reasoning effort levels (low, medium, high) to balance reasoning depth, latency, and cost. It is well suited for autonomous agents, software engineering, multimodal analysis, and large-scale workflow automation.NVIDIA Nemotron 3.5 Super VL is an open multimodal Mixture-of-Experts (MoE) model featuring 120B total parameters with 12B activated per token. Built on a hybrid Mamba–Transformer architecture with multi-token prediction, it supports text, images, documents, GUI screenshots, and video inputs, along with a 1M-token context window for complex, long-horizon workflows. Optimized for multimodal reasoning, visual understanding, and agentic tool use, the model supports configurable reasoning modes and token budgets. It is well suited for document intelligence, web research, coding validation, customer service, GUI analysis, and video understanding, enabling efficient, long-running multimodal agent applications.