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 | Gemini 4 Carbongemini-4-carbon | Qwen3 30B A3B Thinkqwen3-30b-a3b-think | DeepSeek V4.1 Flash 1017deepseek-v4.1-flash-1017 |
|---|---|---|---|
| Pricing | |||
| Input | $2.00 / 1M | $0.051 / 1M | $0.20 / 1M |
| Output | $10.00 / 1M | $0.34 / 1M | $0.80 / 1M |
| Cache Write (5m) | $2.00 / 1M | $0.051 / 1M | $0.20 / 1M |
| Cache Write (1h) | $2.00 / 1M | $0.051 / 1M | $0.20 / 1M |
| Cache Read | $2.00 / 1M | $0.051 / 1M | $0.20 / 1M |
| Web Search | $0 / 1M | $0 / 1M | $0 / 1M |
| Context | |||
| Max context | 1M | 32.8K | 1M |
| Max output | N/A | N/A | N/A |
| Capabilities | |||
| Vision | Yes | No | Yes |
| Function Calling | Yes | Yes | Yes |
| JSON Mode | Yes | Yes | Yes |
| Streaming | Yes | Yes | Yes |
| Catalogue | |||
| Provider | Alibaba | DeepSeek | |
| Category | chat | chat | chat |
| Charge type | Pay As You Go | Pay As You Go | Pay As You Go |
| Released | — | — | — |
| Description | |||
| Summary | 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. | 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. | DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek’s cost-efficient Mixture-of-Experts (MoE) model, built on its innovative Causal Encoder-Decoder (CED) architecture. Featuring 552B total parameters with 8B activated for input and 16B for output, it combines efficient inference with native multimodal understanding through jointly trained text and visual representations. Optimized for coding, terminal and computer-use agents, and long-horizon workflows, the model introduces compressed KV caching that reduces cache memory usage to approximately one-quarter of the previous Flash generation. DeepSeek reports that it surpasses V4 Pro in performance, inference speed, and task completion time, making it well suited for cost-efficient autonomous agents, complex software engineering, and long-context analysis. |