Connic Model Catalog

Exact model IDs, supported inputs, context limits, and EUR token prices for Connic-managed inference.

Configuration

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: anthropic/claude-sonnet-4-6
temperature: 0.3
reasoning_effort: high

Available models

Prices are net EUR per one million tokens. Input covers uncached input and cache creation; cached input applies only to cache hits reported by the supplier. A dash means no discounted cache tariff is published. 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 execution profile and any narrower limits.

A warning label beside a model name reports a property of that model's upstream provider: Non-EU-native provider marks a model whose provider is not EU-native, and 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.

Fast IDs

An ID ending in -fast resolves to a latency-optimized execution path for that model. Most are a second ID beside a standard one and are charged at the same price; a few models are offered only on this path, so -fast is part of their only ID. An ID without the suffix is never served this way.

model: connic/glm-5.2-fast
reasoning_effort: high   # -fast accepts a narrower set than connic/glm-5.2

• A Fast ID may be served by a quantized deployment of the model. That is part of what reduces latency, and it is why outputs may differ from the standard ID on the same prompt. Quantization is a quality trade-off: treat a switch as a model change, re-run the agent's tests, and prefer the standard ID for accuracy-critical steps until an A/B comparison shows the Fast ID holds up on your traffic.
• A Fast ID can be narrower than its standard ID in three ways: fewer reasoning_effort values, a smaller context window, or fewer input modalities. Setting a value the Fast ID does not accept is rejected.
• Not every Fast ID is narrower. Where a standard ID exists, the info icon beside the Fast one lists the differences, and shows none when the two match.

Routing and failover

Multiple inference providers sit behind the catalog. A connic/* ID always resolves to the model named in the catalog; which provider serves a given request is managed by Connic and can change between requests.

• Within the ID. When a route is unavailable and retrying is safe, the request moves to other eligible capacity for the same model. The model identity does not change.
• Uncertain outcomes are not replayed. If a managed request fails in a way that leaves its outcome unknown, Connic neither retries it nor switches to fallback_model. The run surfaces the error instead of risking a duplicate completion.
• Then fallback_model.On a definitive failure, the agent's fallback_model takes over. It can be a managed or a BYOK model, in either direction. A fallback-eligible error switches model immediately rather than consuming the remaining retry_options.attempts.
• Runs that switched are marked. Filter the run list with the expression context.fallback_model_used, and see the switch in the run's trace.

Retries apply to the individual model request, not the whole run: tool calls that already executed keep their results. Configure attempts and backoff under execution limits and retries.

Cancelled requests

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.

EU inference and data handling

• Every connic/* request is sent only to EU inference capacity. If that boundary cannot be maintained for a model, the model is removed from the catalog or the call fails closed. It is never rerouted outside the EU.
• Managed inference is not used to train models. Connic records model and token counts for billing rather than prompt or response content; narrow security and error-retention exceptions are described in the Privacy Policy.

EU Project boundary

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.