Exact model IDs, supported inputs, context limits, and EUR token prices for Connic-managed inference.
On prepaid Projects, enable Auto-refill under Project → Billing before sending production traffic to connic/* models. If available credit runs out, new Project usage pauses until credit is added.
Use an exact catalog ID as the primary or fallback model. A Project can mix Connic-managed and BYOK models.
model: connic/glm-5.2
fallback_model: connic/gpt-5.6-luna # or anthropic/claude-sonnet-4-6 with Anthropic BYOK configured
temperature: 0.3
reasoning_effort: highPrices are net EUR per one million tokens. A dash means no cached-input discount is available. Listed reasoning effort values are the model-specific YAML overrides. Omit reasoning_effort or use auto for model-managed behavior. Standard and Fast IDs share one row and the same prices. Open the info icon beside a Fast ID for its supported reasoning settings, context limit, and input modalities.
| Capabilities | ||||
|---|---|---|---|---|
Non-EU-native provider Claude Haiku 4.5 connic/claude-haiku-4-5Fast model for high-volume coding and agent tasks. Anthropic · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 200k | €1.10 €0.15 (cached) | €5.10 |
Non-EU-native provider Claude Opus 5 connic/claude-opus-5Advanced model for long-running agents, complex coding, and professional work. Anthropic · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €5.20 €0.55 (cached) | €26.00 |
Non-EU-native provider Claude Sonnet 5 connic/claude-sonnet-5High-performance model for coding, agents, and everyday professional tasks. Anthropic · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €2.10 €0.25 (cached) | €10.40 |
EU-native provider Codestral 2508 connic/codestral-2508Low-latency coding model for completion, correction, and test generation. Mistral · Provider terms Reasoning effort Not supported | Text | 256k | €0.35 €0.05 (cached) | €0.95 |
EU-native provider DeepSeek V4 Flash 0731 connic/deepseek-v4-flash-0731connic/deepseek-v4-flash-0731-fastUpdated open-weight reasoning model for long-context coding and agentic work. DeepSeek · MIT Reasoning effort none, low, high, max | TextReasoning | 256k | €0.45 €0.10 (cached) | €0.85 |
EU-native provider DeepSeek V4 Pro connic/deepseek-v4-pro-fastFrontier open-weight reasoning model for complex long-context coding and agentic work. DeepSeek · MIT Reasoning effort high, max | TextReasoning | 1M | €1.70 €0.45 (cached) | €3.30 |
Non-EU-native provider Gemini 3.1 Flash Lite connic/gemini-3.1-flash-liteLow-latency multimodal model for cost-sensitive, high-volume workloads. Google · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €0.30 €0.05 (cached) | €1.60 |
Non-EU-native provider Gemini 3.5 Flash connic/gemini-3.5-flashFast frontier multimodal model for agents, coding, and large-scale tasks. Google · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €1.60 €0.20 (cached) | €9.40 |
Non-EU-native provider Gemini 3.5 Flash Lite connic/gemini-3.5-flash-liteLow-latency multimodal model for high-volume subagent workflows and document parsing. Google · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €0.35 €0.05 (cached) | €2.60 |
Non-EU-native provider Gemini 3.6 Flash connic/gemini-3.6-flashFast multimodal model for agentic coding, long-context reasoning, and multi-step workflows. Google · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €0.75 €0.10 (cached) | €3.60 |
Non-EU-native provider Gemini 3.7 Flash connic/gemini-3.7-flashFast multimodal model for agentic coding, long-context reasoning, and multi-step workflows. Google · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €0.75 €0.10 (cached) | €3.60 |
Non-EU-native provider Gemini 3.8 Flash connic/gemini-3.8-flashFast multimodal model for coding, agents, and long-context reasoning. Google · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €0.80 €0.10 (cached) | €3.90 |
EU-native provider Gemma 4 26B A4B connic/gemma-4-26b-a4b-itEfficient open-weight agentic and reasoning model with image understanding. Google · Apache-2.0 Reasoning effort none, low, medium, high | TextVisionReasoning | 256k | €0.30 | €0.55 |
EU-native provider Gemma 4 31B IT connic/gemma-4-31b-itMultimodal instruction model for reasoning, coding, and vision-language tasks. Google · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 128k | €0.30 | €0.55 |
EU-native provider GLM 5.2 connic/glm-5.2connic/glm-5.2-fastOpen-weight reasoning model for long-horizon agentic work. Z.ai · MIT Reasoning effort none, high, max | TextReasoning | 256k | €1.90 €0.40 (cached) | €5.80 |
EU-native provider GPT OSS 120B connic/gpt-oss-120bconnic/gpt-oss-120b-fastLarge open-weight mixture-of-experts model for reasoning, agents, and developer workloads. OpenAI · Provider terms Reasoning effort low, medium, high | TextReasoning | 128k | €0.25 | €0.70 |
Non-EU-native provider30-day provider retention GPT-5.4 connic/gpt-5.4Frontier multimodal model for professional workflows, coding, and agents. OpenAI · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €2.80 €0.25 (cached) | €14.60 |
Non-EU-native provider30-day provider retention GPT-5.6 Luna connic/gpt-5.6-lunaFast multimodal model for high-volume work and lightweight reasoning. OpenAI · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €0.25 €0.05 (cached) | €1.30 |
Non-EU-native provider30-day provider retention GPT-5.6 Sol connic/gpt-5.6-solFlagship multimodal model for advanced reasoning, deep research, and complex multi-step execution. OpenAI · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €5.20 €0.55 (cached) | €31.20 |
Non-EU-native provider30-day provider retention GPT-5.6 Terra connic/gpt-5.6-terraBalanced multimodal model for professional work, coding, and reasoning. OpenAI · Provider terms Reasoning effort low, medium, high | TextVisionReasoning | 1M | €2.10 €0.25 (cached) | €12.50 |
EU-native provider Kimi K2.7 Code connic/kimi-k2.7-code-fastOpen-weight agentic coding model for long-horizon software engineering. Moonshot AI · Modified MIT Reasoning effort Automatic (not configurable) | TextVisionReasoning | 256k | €1.20 €0.30 (cached) | €4.20 |
EU-native provider Kimi K3 connic/kimi-k3-fastFrontier open-weight multimodal model for long-horizon coding, reasoning, and agentic work. Moonshot AI · Kimi K3 License Reasoning effort Automatic (not configurable) | TextVisionReasoning | 1M | €2.90 | €14.20 |
EU-native provider Llama 3.3 70B Instruct connic/llama-3.3-70b-instructconnic/llama-3.3-70b-instruct-fastEstablished open-weight multilingual instruction model for text workloads. Meta · Llama 3.3 Community Reasoning effort Not supported | Text | 100k | €0.95 | €0.95 |
EU-native provider MiniMax M2.7 connic/minimax-m2.7Autonomous agent model for multi-agent coordination and enterprise-scale work. MiniMax · Provider terms Reasoning effort low, medium, high | TextReasoning | 197k | €0.65 | €2.60 |
EU-native provider Ministral 3 14B connic/ministral-14b-2512Compact frontier-level multimodal model for text, vision, agents, and local deployment. Mistral · Provider terms Reasoning effort Not supported | TextVision | 256k | €0.25 €0.05 (cached) | €0.25 |
EU-native provider Mistral Large 2512 connic/mistral-large-2512Open-weight multimodal mixture-of-experts model for text and vision tasks. Mistral · Provider terms Reasoning effort Not supported | TextVision | 256k | €0.55 €0.10 (cached) | €1.60 |
EU-native provider Mistral Medium 3.5 128B connic/mistral-medium-3.5-128bUnified open-weight model for instruction following, reasoning, and image understanding. Mistral · Modified MIT Reasoning effort none, high | TextVisionReasoning | 180k | €1.60 €0.15 (cached) | €7.90 |
EU-native provider Mistral Small 3.2 24B connic/mistral-small-3.2-24b-instruct-2506Efficient open-weight text and vision model for responsive agent workloads. Mistral · Apache-2.0 Reasoning effort Not supported | TextVision | 128k | €0.15 | €0.35 |
EU-native provider Mistral Small 4 2603 connic/mistral-small-2603connic/mistral-small-2603-fastEfficient multimodal model for automatic reasoning, agents, and coding. Mistral · Provider terms Reasoning effort Automatic (not configurable) | TextVisionReasoning | 256k | €0.25 €0.05 (cached) | €0.65 |
EU-native provider Qwen 3 235B A22B Instruct connic/qwen3-235b-a22b-instruct-2507Large open-weight multilingual instruction model for demanding text workloads. Qwen · Apache-2.0 Reasoning effort Not supported | Text | 250k | €0.75 | €2.30 |
EU-native provider Qwen 3.5 397B A17B connic/qwen3.5-397b-a17bconnic/qwen3.5-397b-a17b-fastFrontier open-weight reasoning model with image understanding. Qwen · Apache-2.0 Reasoning effort none, low, medium, high | TextVisionReasoning | 250k | €0.65 | €3.80 |
EU-native provider Qwen 3.6 35B A3B connic/qwen3.6-35b-a3bconnic/qwen3.6-35b-a3b-fastFast open-weight model for agentic tasks, reasoning, and image understanding. Qwen · Apache-2.0 Reasoning effort none, low, medium, high | TextVisionReasoning | 256k | €0.30 | €1.60 |
EU-native provider Voxtral Small 2507 connic/voxtral-small-2507Efficient Mistral model for multilingual text understanding and generation. Mistral · Provider terms Reasoning effort Not supported | Text | 32k | €0.15 €0.05 (cached) | €0.35 |
connic/claude-haiku-4-5Fast model for high-volume coding and agent tasks.
low, medium, highconnic/claude-opus-5Advanced model for long-running agents, complex coding, and professional work.
low, medium, highconnic/claude-sonnet-5High-performance model for coding, agents, and everyday professional tasks.
low, medium, highconnic/codestral-2508Low-latency coding model for completion, correction, and test generation.
Not supportedconnic/deepseek-v4-flash-0731connic/deepseek-v4-flash-0731-fastUpdated open-weight reasoning model for long-context coding and agentic work.
none, low, high, maxconnic/deepseek-v4-pro-fastFrontier open-weight reasoning model for complex long-context coding and agentic work.
high, maxconnic/gemini-3.1-flash-liteLow-latency multimodal model for cost-sensitive, high-volume workloads.
low, medium, highconnic/gemini-3.5-flashFast frontier multimodal model for agents, coding, and large-scale tasks.
low, medium, highconnic/gemini-3.5-flash-liteLow-latency multimodal model for high-volume subagent workflows and document parsing.
low, medium, highconnic/gemini-3.6-flashFast multimodal model for agentic coding, long-context reasoning, and multi-step workflows.
low, medium, highconnic/gemini-3.7-flashFast multimodal model for agentic coding, long-context reasoning, and multi-step workflows.
low, medium, highconnic/gemini-3.8-flashFast multimodal model for coding, agents, and long-context reasoning.
low, medium, highconnic/gemma-4-26b-a4b-itEfficient open-weight agentic and reasoning model with image understanding.
none, low, medium, highconnic/gemma-4-31b-itMultimodal instruction model for reasoning, coding, and vision-language tasks.
low, medium, highconnic/glm-5.2connic/glm-5.2-fastOpen-weight reasoning model for long-horizon agentic work.
none, high, maxconnic/gpt-oss-120bconnic/gpt-oss-120b-fastLarge open-weight mixture-of-experts model for reasoning, agents, and developer workloads.
low, medium, highconnic/gpt-5.4Frontier multimodal model for professional workflows, coding, and agents.
low, medium, highconnic/gpt-5.6-lunaFast multimodal model for high-volume work and lightweight reasoning.
low, medium, highconnic/gpt-5.6-solFlagship multimodal model for advanced reasoning, deep research, and complex multi-step execution.
low, medium, highconnic/gpt-5.6-terraBalanced multimodal model for professional work, coding, and reasoning.
low, medium, highconnic/kimi-k2.7-code-fastOpen-weight agentic coding model for long-horizon software engineering.
Automatic (not configurable)connic/kimi-k3-fastFrontier open-weight multimodal model for long-horizon coding, reasoning, and agentic work.
Automatic (not configurable)connic/llama-3.3-70b-instructconnic/llama-3.3-70b-instruct-fastEstablished open-weight multilingual instruction model for text workloads.
Not supportedconnic/minimax-m2.7Autonomous agent model for multi-agent coordination and enterprise-scale work.
low, medium, highconnic/ministral-14b-2512Compact frontier-level multimodal model for text, vision, agents, and local deployment.
Not supportedconnic/mistral-large-2512Open-weight multimodal mixture-of-experts model for text and vision tasks.
Not supportedconnic/mistral-medium-3.5-128bUnified open-weight model for instruction following, reasoning, and image understanding.
none, highconnic/mistral-small-3.2-24b-instruct-2506Efficient open-weight text and vision model for responsive agent workloads.
Not supportedconnic/mistral-small-2603connic/mistral-small-2603-fastEfficient multimodal model for automatic reasoning, agents, and coding.
Automatic (not configurable)connic/qwen3-235b-a22b-instruct-2507Large open-weight multilingual instruction model for demanding text workloads.
Not supportedconnic/qwen3.5-397b-a17bconnic/qwen3.5-397b-a17b-fastFrontier open-weight reasoning model with image understanding.
none, low, medium, highconnic/qwen3.6-35b-a3bconnic/qwen3.6-35b-a3b-fastFast open-weight model for agentic tasks, reasoning, and image understanding.
none, low, medium, highconnic/voxtral-small-2507Efficient Mistral model for multilingual text understanding and generation.
Not supportedProvider labels beside a model name report whether the provider is EU-native. EU-native provider marks an EU-native provider, while Non-EU-native provider marks a model whose provider is not EU-native. The additional 30-day provider retention marks one whose provider retains request data for 30 days under its own policy. Inference still runs on EU capacity in both cases.
An ID ending in -fast selects a latency-optimized model variant. Most models list standard and Fast IDs at the same price; some list only a Fast ID.
model: connic/glm-5.2-fast
reasoning_effort: high # -fast accepts a narrower set than connic/glm-5.2reasoning_effort values, a smaller context window, or fewer input modalities. Setting a value the Fast ID does not accept is rejected.Model request failures use retry_options. Without a fallback, the primary uses the configured attempt budget. With a configured fallback_model, the primary is tried once and the fallback uses that budget. A switch marks the run with context.fallback_model_used.
Retries apply to the individual model request, not the whole run, so completed tool calls keep their results. Configure attempts and backoff under execution limits and retries.
A managed request that ends early, because the run hit its timeout or the run was stopped, is settled at zero tokens and the credit reserved for it is released back to the Project balance. This applies to streaming and non-streaming requests alike.
connic/* request runs in the EU. A request fails when EU execution is unavailable.The connic/* guarantee covers the model call. A complete Project remains EU-only when every customer-configured component, including BYOK providers, tools, guardrails, judges, and external data destinations, also stays in the EU.