Skip to content
Seedance 2.0 Face is here — generate video from real-person reference photos.Try it now →

MODEL MARKETPLACE

Explore All Models

Browse mainstream AI models from all major providers. Compare pricing, capabilities, and find the perfect model for your use case.

All available models

Browse 30 selected AI models on ElliotGate. Each link opens the model detail page with pricing, capabilities, and integration examples.

Image 0 Video 0
30 models found
OOpenAI: GPT-5.5openai/gpt-5.5GPT-5.5 is OpenAI's frontier model, a fully retrained base that builds on GPT-5.4 with stronger reasoning, higher reliability, and improved token efficiency on hard tasks. It exposes multiple reasoning effort levels (from non-reasoning to xhigh) and accepts text and image inputs within a 1M+ token context window (922K input, 128K output). Reported benchmarks include 82.7% on Terminal-Bench 2.0, 51.7% on FrontierMath Tier 1-3, 35.4% on FrontierMath Tier 4, and 84.9% on GDPval. It is aimed at complex professional, coding, and multimodal workflows that require large-scale reasoning in a single system.
OOpenAI: GPT-5.5 Proopenai/gpt-5.5-proGPT-5.5 Pro is a high-capability model from OpenAI optimized for deep reasoning and accuracy on complex workloads. It provides a context window of roughly 1,050,000 tokens (922K input, 128K output) and accepts text and image input. The model is designed for long-horizon problem solving, agentic coding, and multi-step execution, and exposes selectable reasoning effort levels. It does not offer a cached input discount.
OOpenAI: GPT-5.4openai/gpt-5.4GPT-5.4 is OpenAI's frontier model that unifies the Codex and GPT lines into a single system. It provides a 1M+ token context window (roughly 922K input, 128K output) and accepts text and image inputs, enabling high-context reasoning, coding, and multimodal analysis in one workflow. The model adds native computer use and tool search, reporting 75.0% on OSWorld-Verified and 57.7% on SWE-Bench Pro; tool search loads tool definitions on demand and reduced total token usage by 47% in MCP testing. It is positioned as a strong default for both general-purpose tasks and software engineering. For inputs above 272K tokens, pricing applies a higher per-token rate for the full session.
OOpenAI: GPT-5.4 Miniopenai/gpt-5.4-miniGPT-5.4 Mini brings the core capabilities of GPT-5.4 to a faster, more efficient model optimized for high-throughput workloads, running more than 2x faster than its predecessor while lowering latency and cost. It accepts text, image, and file inputs with performance across reasoning, coding, and tool use, making it suited to chat applications, coding assistants, and agent workflows that operate at scale. The model provides a 400K-token context window with up to 128K tokens of output and supports reasoning. GPT-5.4 Mini was released in March 2026.
OOpenAI: GPT-5openai/gpt-5GPT-5 is OpenAI's flagship model, generally available since August 2025, with improvements in reasoning, code quality, and instruction following. It is optimized for complex tasks requiring step-by-step reasoning and accuracy in high-stakes use cases, and supports user-specified intent such as asking the model to think harder. On coding benchmarks it scores 74.9% on SWE-bench Verified and 88% on Aider Polyglot. It accepts text, image, and file inputs and returns text, with a 400K-token context window and up to 128K output tokens. OpenAI reports reductions in hallucination and sycophancy relative to prior models.
OOpenAI: GPT-5 Miniopenai/gpt-5-miniGPT-5 Mini is a compact variant of GPT-5 built for lighter-weight reasoning tasks at reduced latency and cost. It retains the instruction-following and safety-tuning characteristics of GPT-5 and is positioned as the successor to OpenAI's o4-mini. It accepts text, image, and file inputs and returns text, with a 400K-token context window and up to 128K output tokens. The API supports prompt caching and an optional web search tool.
OOpenAI: GPT-4o (extended)openai/gpt-4oGPT-4o ("o" for "omni") is an OpenAI model that accepts text and image inputs and returns text outputs, with a 128K context window. It matches GPT-4 Turbo on English text while improving non-English language handling and visual understanding. On MMLU it scores around 88.7% (0-shot CoT), and it reaches strong results on vision benchmarks such as MMMU, MathVista, and ChartQA. Access is provided through OpenAI-compatible API endpoints with support for function calling and JSON mode.
OOpenAI: GPT-4o-miniopenai/gpt-4o-miniGPT-4o mini is OpenAI's small multimodal model, accepting text and image inputs and producing text outputs over a 128K token context window with up to 16,384 output tokens. It is positioned as a low-cost option below the GPT-4o line, with cached input pricing available. OpenAI reports an MMLU score of 82% for the model. It supports tool use and structured outputs through the API.
AAnthropic: Claude Opus 4.8anthropic/claude-opus-4.8Claude Opus 4.8 is Anthropic's most capable generally available model in the Opus family, released in May 2026. It accepts text, image, and file inputs with text output, supports reasoning, and provides a 1M-token context window with up to 128K output tokens. It is suited for highly autonomous agents, long-horizon agentic work, complex coding, and knowledge work where coherence over extended sessions matters, including large codebases, multi-stage debugging, document drafting, and data analysis. On reported benchmarks it reaches 69.2% on SWE-bench Pro, up from 64.3% for Opus 4.7, and 88.6% on SWE-bench Verified. The release also emphasizes higher honesty, with the model roughly four times less likely than its predecessor to let flaws in its own code pass unremarked.
AAnthropic: Claude Opus 4.7anthropic/claude-opus-4.7Claude Opus 4.7 is Anthropic's next-generation Opus model, built for long-running, asynchronous agents and complex multi-step workflows. Building on Opus 4.6, it improves software engineering on the hardest tasks, adds configurable task budgets and an xhigh effort level, and follows instructions more strictly with fewer tool errors. It accepts text and image input with a 1M token context window and up to 128K output tokens, and supports reasoning. Vision handling is also stronger, including higher-resolution image input. Beyond coding, it targets knowledge work such as drafting documents, building presentations, and analyzing data while maintaining coherence across extended sessions.
AAnthropic: Claude Sonnet 4.6anthropic/claude-sonnet-4.6Claude Sonnet 4.6 is Anthropic's most capable Sonnet-class model, with strong performance across coding, agents, and professional work. It scores 79.6% on SWE-bench Verified, 59.1% on Terminal-Bench 2.0, and 72.5% on OSWorld-Verified, all improvements over Sonnet 4.5. It excels at iterative development, complex codebase navigation, end-to-end project management with memory, polished document creation, and computer use for web QA and workflow automation. Anthropic reports improved resistance to prompt injection over Sonnet 4.5. The model supports a 1M-token context window with up to 128K tokens of output, and accepts text and image input.
AAnthropic: Claude Sonnet 4.5anthropic/claude-sonnet-4.5Claude Sonnet 4.5 is Anthropic's Sonnet model optimized for real-world agents and coding workflows. Anthropic reports 77.2% on SWE-bench Verified for software coding and 61.4% on OSWorld for computer use, with improvements across system design, code security, and specification adherence. The model is designed for extended autonomous operation, maintaining task continuity across sessions, and adds stronger agentic capabilities including improved tool orchestration, speculative parallel execution, and more efficient context and memory management. It accepts text, image, and file input, produces text, and supports a 1,000,000-token context window.
AAnthropic: Claude Haiku 4.5anthropic/claude-haiku-4.5Claude Haiku 4.5 is Anthropic's fastest and most cost-efficient model, designed to deliver near-frontier intelligence at lower latency and cost than larger Claude models. It matches Claude Sonnet 4 on reasoning, coding, and computer-use tasks while running at roughly twice the speed and one-third the cost. It is the first Haiku model to add extended thinking, with controllable reasoning depth and summarized or interleaved thought output, alongside support for tool use, bash, web search, and computer use. Haiku 4.5 scores 73.3% on SWE-bench Verified and supports a 200K token context window with up to 64K output tokens for text and image inputs.
GGoogle: Gemini 3.1 Pro Previewgoogle/gemini-3.1-pro-previewGemini 3.1 Pro Preview is Google's frontier reasoning model, delivering enhanced software engineering performance, improved agentic reliability, and more efficient token usage across complex workflows. Building on the multimodal foundation of the Gemini 3 series, it combines reasoning across text, image, video, audio, and code with a 1M-token context window. The 3.1 update introduces gains in SWE benchmarks and real-world coding environments, along with stronger autonomous task execution in structured domains such as finance and spreadsheet-based workflows. It introduces a new medium thinking level to balance cost, speed, and performance, and is suited to agentic coding, structured planning, multimodal analysis, and workflow automation.
GGoogle: Gemini 3.1 Flash Litegoogle/gemini-3.1-flash-liteGemini 3.1 Flash Lite is Google's most cost-efficient Gemini model, generally available and optimized for low-latency, high-volume workloads. It accepts text, image, video, audio, and PDF inputs with text output, supports a roughly 1M-token context window with up to 64K output tokens, and targets lightweight agentic workflows, simple data extraction, and applications where responsiveness and API cost are the primary constraints. It supports configurable thinking levels (minimal, low, medium, high) for fine-grained cost/performance trade-offs and includes targeted improvements for instruction following and audio-input quality.
GGoogle: Gemini 3.5 Flashgoogle/gemini-3.5-flashGemini 3.5 Flash is Google's high-efficiency multimodal model, delivering coding and reasoning quality at Flash-tier cost and speed. It accepts text, image, video, audio, and PDF inputs with text output, supports a roughly 1M-token context window with up to 64K output tokens, and is optimized for coding and parallel agentic execution loops. On benchmarks Google reports, it scores 76.2% on Terminal-Bench 2.1 and 83.6% on MCP Atlas. It defaults to medium thinking effort and supports configurable thinking levels (minimal, low, medium, high) for cost/performance trade-offs.
DDeepSeek: DeepSeek V4 Prodeepseek/deepseek-v4-proDeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. Each MoE layer uses 384 routed experts plus 1 shared expert, with 6 experts active per token. It combines a hybrid attention mechanism (Compressed Sparse Attention and Heavily Compressed Attention) for efficient long-context processing and supports multiple reasoning modes to balance speed and depth. Reported benchmarks include 80.6% on SWE-bench Verified, 67.9% on Terminal-Bench 2.0, and 93.5% on LiveCodeBench. It is suited to full-codebase analysis, multi-step automation, and large-scale information synthesis.
DDeepSeek: DeepSeek V4 Flashdeepseek/deepseek-v4-flashDeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window. It shares the V4 architecture with DeepSeek V4 Pro, including a hybrid attention mechanism (Compressed Sparse Attention and Heavily Compressed Attention) for efficient long-context processing and configurable reasoning modes. The design targets fast inference and high-throughput workloads while maintaining reasoning and coding performance, making it suitable for coding assistants, chat systems, and agent workflows where responsiveness and cost efficiency matter.
DDeepSeek: R1deepseek/deepseek-r1DeepSeek R1 is an open-source reasoning model from DeepSeek with performance on par with OpenAI o1, released with fully open reasoning tokens. It is a Mixture-of-Experts (MoE) model with 671B total parameters and 37B active per inference pass. R1 is trained with large-scale reinforcement learning to reason via chain of thought, and reports 79.8% on AIME 2024, 97.3% on MATH-500, and a 96.3 percentile ranking on Codeforces. The model is released under an MIT license, permitting distillation and commercial use.
DDeepSeek: DeepSeek V3deepseek/deepseek-chatDeepSeek-V3 is a Mixture-of-Experts (MoE) model from the DeepSeek team with 671B total parameters and 37B activated per token, building on the instruction-following and coding abilities of prior versions. It adopts Multi-head Latent Attention (MLA) and the DeepSeekMoE architecture, with an auxiliary-loss-free load-balancing strategy and a multi-token prediction objective. Pre-trained on nearly 15 trillion tokens, it reports 87.1% on MMLU, 89.3% on GSM8K, and 65.2% on HumanEval, with results that rival leading closed-source models.
QQwen: Qwen3.6 Max Previewqwen/qwen3.6-max-previewQwen3.6-Max-Preview is a proprietary frontier model from Alibaba built on a sparse mixture-of-experts (MoE) architecture with approximately 1 trillion total parameters. It is optimized for agentic coding, tool use, and long-context reasoning, with a 262K token context window. The model reports top scores across several coding benchmarks, including SWE-bench Pro, Terminal-Bench 2.0, and SciCode, and an AA Intelligence Index score of 52. It includes an integrated thinking mode that preserves reasoning traces across multi-turn conversations and supports structured output and function calling. No open weights are provided.
QQwen: Qwen3.6 Plusqwen/qwen3.6-plusQwen3.6 Plus is Alibaba's flagship model built on a hybrid architecture that combines efficient linear attention with sparse Mixture-of-Experts (MoE) routing for scalable, high-performance inference. Compared with the 3.5 series, it delivers gains in agentic coding, front-end development, and overall reasoning, and handles complex tasks such as 3D scenes, games, and repository-level problem solving. It scores 78.8 on SWE-bench Verified. The model accepts text, image, and video input with text output, supports reasoning, and provides a 1M token context window with up to 65K output tokens.
QQwen: Qwen3.6 Flashqwen/qwen3.6-flashQwen3.6 Flash is the speed-first variant in Alibaba's Qwen 3.6 series. It accepts text, image, and video input with a 1M token context window and includes an integrated thinking mode. The model supports tool use and structured output, and prompt caching is available with separate cache read and cache write pricing. Pricing is tiered, with higher rates applied to requests above the 256K token threshold.
XxAI: Grok 4.3x-ai/grok-4.3Grok 4.3 is a reasoning model from xAI, released in 2026 as an incremental upgrade focused on reasoning quality, response speed, and tool-use reliability. It accepts text and image inputs with text output and is suited for agentic workflows, instruction-following tasks, and applications requiring high factual accuracy. Reasoning effort can be configured across none, low, medium, and high levels, with low as the default. It supports a 1M-token context window with no fixed output token limit, making it applicable to long-document analysis, deep research, and multi-step agentic tasks. Pricing is tiered: requests exceeding 200K total tokens are billed at a higher rate.
XxAI: Grok 4.20x-ai/grok-4.20Grok 4.20 is xAI's flagship model, emphasizing inference speed and agentic tool calling. xAI positions it for strict prompt adherence and a low hallucination rate; on the Artificial Analysis Omniscience test it reached a 78% non-hallucination rate, the highest among models tested, and ranked second on tau-squared-Bench Telecom with 97% for agentic tool use. Reasoning can be enabled or disabled via the reasoning enabled parameter in the API. The model accepts text, image, and file inputs and supports a 2,000,000-token context window.
XxAI: Grok Code Fast 1x-ai/grok-code-fast-1Grok Code Fast 1 is a low-cost reasoning model from xAI built specifically for agentic coding workflows, with reasoning traces visible in the response so developers can steer its behavior. It was built from a separate architecture with a pre-training corpus of programming-related data and post-training drawn from real pull requests and practical coding tasks, and is optimized for tool use such as grep, terminal operations, and file editing across languages including TypeScript, Python, Java, Rust, C++, and Go. The model emphasizes speed and prompt caching for high-volume iterative work, supports a 256,000-token context window, and reportedly scored 70.8% on the SWE-Bench-Verified subset using xAI's internal harness.
MMeta: Llama 4 Maverickmeta-llama/llama-4-maverickLlama 4 Maverick is a multimodal model from Meta built on a mixture-of-experts (MoE) architecture with 128 routed experts and a shared expert, activating 17 billion parameters per forward pass out of 400B total. It accepts text and image input and produces text and code output across 12 supported languages, using early fusion for native multimodality and a 1 million token context window. Trained on roughly 22 trillion tokens with a knowledge cutoff in August 2024, it is instruction-tuned for assistant-style interaction, image reasoning, and tool use. On the LMArena chat leaderboard it reached an ELO score of 1417, and it performs comparably to DeepSeek V3 on reasoning and coding with fewer active parameters. Released April 5, 2025 under the Llama 4 Community License.
FFLUX.2 Maxblack-forest-labs/flux-2-maxA selected ElliotGate model with pricing, capabilities, and API access through a single gateway.
DDreamina Seedance 2.0bytedance/dreamina-seedance-2-0-260128A selected ElliotGate model with pricing, capabilities, and API access through a single gateway.
MMiniMax: MiniMax M3minimax/minimax-m3MiniMax-M3 is a natively multimodal Mixture-of-Experts foundation model from MiniMax with 229.9B total parameters, activating about 9.8B per token across 256 fine-grained experts. It supports text, image, and video inputs with text output, a 1M-token context window, and up to 512K output tokens, and is suited for long-horizon agentic work, coding, and tool use. It is built on MiniMax Sparse Attention (MSA), which uses a lightweight index branch to select relevant KV-cache blocks so the main attention layer processes only those blocks, cutting per-token compute at 1M context to roughly 1/20 of the previous M2 generation. MSA delivers about 9.7x faster prefill and 15.6x faster decode at 1M tokens versus M2 while retaining accuracy, and the model reaches 59.0% on SWE-bench Pro.