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 | Muse Spark Openmuse-spark-open | Nemotron 3.5 Super VLnemotron-3.5-super-vl | Qwen3 30B A3B Thinkqwen3-30b-a3b-think |
|---|---|---|---|
| Pricing | |||
| Input | $1.25 / 1M | $0.10 / 1M | $0.051 / 1M |
| Output | $4.25 / 1M | $0.40 / 1M | $0.34 / 1M |
| Cache Write (5m) | $1.25 / 1M | $0.10 / 1M | $0.051 / 1M |
| Cache Write (1h) | $1.25 / 1M | $0.10 / 1M | $0.051 / 1M |
| Cache Read | $1.25 / 1M | $0.10 / 1M | $0.051 / 1M |
| Web Search | $0 / 1M | $0 / 1M | $0 / 1M |
| Context | |||
| Max context | 1M | 1M | 32.8K |
| Max output | N/A | N/A | N/A |
| Capabilities | |||
| Vision | Yes | Yes | No |
| Function Calling | Yes | Yes | Yes |
| JSON Mode | Yes | Yes | Yes |
| Streaming | Yes | Yes | Yes |
| Catalogue | |||
| Provider | Meta | NVIDIA | Alibaba |
| Category | chat | chat | chat |
| Charge type | Pay As You Go | Pay As You Go | Pay As You Go |
| Released | — | — | — |
| Description | |||
| Summary | Muse Spark Open is an open-weight, coding-optimized model designed for agentic software engineering workflows. It delivers strong first-attempt accuracy, reliable tool calling, and deep codebase understanding, enabling effective development across real-world repositories. Optimized for autonomous coding agents, complex code navigation, and tool-driven development, Muse Spark Open is well suited for software engineering automation, debugging, code generation, and multi-step coding workflows where accuracy and reliable execution are essential. | 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. | Qwen3-30B-A3B-Thinking-2507 is a 30B-parameter Mixture-of-Experts (MoE) reasoning model optimized for complex tasks that require extended, multi-step reasoning. It is purpose-built for thinking mode, where internal reasoning traces are explicitly separated from final outputs, enabling more structured and reliable problem solving. Compared to earlier Qwen3-30B variants, this release delivers notable gains across logical reasoning, mathematics, science, coding, and multilingual benchmarks, while also strengthening instruction following, tool usage, and alignment with human preferences. With improved reasoning efficiency and larger output budgets, Qwen3-30B-A3B-Thinking-2507 is well suited for advanced research, competitive problem solving, and agentic applications that demand robust long-context and structured reasoning capabilities. |