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.
| Attribute | Nemotron 3.5 Super VLnemotron-3.5-super-vl | Kimi K2 Thinkingkimi-k2-thinking |
|---|---|---|
| Pricing | ||
| Input | $0.10 / 1M | $0.34 / 1M |
| Output | $0.40 / 1M | $1.50 / 1M |
| Cache Write (5m) | $0.10 / 1M | $0.34 / 1M |
| Cache Write (1h) | $0.10 / 1M | $0.34 / 1M |
| Cache Read | $0.10 / 1M | $0.34 / 1M |
| Web Search | $0 / 1M | $0 / 1M |
| Context | ||
| Max context | 1M | 262.1K |
| Max output | N/A | N/A |
| Capabilities | ||
| Vision | Yes | Yes |
| Function Calling | Yes | Yes |
| JSON Mode | Yes | Yes |
| Streaming | Yes | Yes |
| Catalogue | ||
| Provider | NVIDIA | Moonshot AI |
| Category | chat | chat |
| Charge type | Pay As You Go | Pay As You Go |
| Released | — | — |
| Description | ||
| Summary | 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. | Kimi K2 Thinking is Moonshot AI's most advanced open reasoning model, built on a trillion-parameter MoE with 32B active parameters and a 256K context window. It's optimized for long-horizon, step-by-step reasoning with dynamic tool use, enabling sustained autonomous research, coding, and writing over hundreds of turns. It sets new open-source records on benchmarks like HLE and LiveCodeBench, and maintains stable multi-agent, tool-heavy workflows (200–300 calls) while balancing deep reasoning with efficient inference. |