Fim Completions Beta
DeepSeek API · DeepSeek
Foundation Model & API — DeepSeek
Evaluates DeepSeek's FIM / Completions (beta) across 9 scenario-based test cases, each graded against an expected-behavior rubric by an LLM judge, from Corsac's Foundation Model & API eval coverage.
About DeepSeek
DeepSeek is an AI company shipping frontier open-weight models (DeepSeek-V3, DeepSeek-R1) and an OpenAI-compatible API with a separate reasoner model (deepseek-reasoner), automatic disk-based context caching, function calling, JSON output, and very low token pricing. The models are released under an MIT license alongside the hosted API.
Sample tests· showing 3 of 9
| # | Input | Expected behavior | Check |
|---|---|---|---|
| 01 | An integrator calls the FIM/completions endpoint against the standard base_url https://api.deepseek.com and gets a 404. | The FIM / completions feature is a beta endpoint — call it with base_url https://api.deepseek.com/beta (the documented beta host). Do not route beta completions through the standard chat host. | Pass / FailAi Platformhigh |
| 02 | A FIM completion for a single function keeps generating into the next function because no stop sequence is set. | Set stop tokens appropriate to the language/context (e.g. a closing brace at the right indentation, or a newline-delimited boundary) so the completion ends at the intended span. Cap max_tokens as a backstop. | Pass / FailAi Platformmedium |
| 03 | A code-completion feature wants the model to fill a gap between existing before-text and after-text, but sends only the prompt (before) with no suffix. | Use the FIM shape: send the leading code as prompt and the trailing code as suffix so the model completes the middle. Omitting suffix turns FIM into a plain prefix completion and the model cannot honor the following context. | Pass / FailAi Platformhigh |
How this eval is graded
Grade against expected.ideal_behavior and expected.rubric. Per-criterion pass requires mean >= 4.0 and no criterion below 3.
Rubric criteria
- Deepseek
- Ai Platform
- Fim Completions Beta
Recommended for
Works with
Related evals
Claude API
Evaluates Anthropic's Batch API across 9 scenario-based test cases, each graded against an expected-behavior rubric by an LLM judge, from Corsac's Foundation Model & API eval coverage.
View AI PlatformClaude API
Evaluates Anthropic's Extended Thinking across 9 scenario-based test cases, each graded against an expected-behavior rubric by an LLM judge, from Corsac's Foundation Model & API eval coverage.
View AI PlatformClaude API
Evaluates Anthropic's Files API & Citations across 9 scenario-based test cases, each graded against an expected-behavior rubric by an LLM judge, from Corsac's Foundation Model & API eval coverage.
ViewFrequently asked questions
What does the Fim Completions Beta eval for DeepSeek DeepSeek API test?+
Evaluates DeepSeek's FIM / Completions (beta) across 9 scenario-based test cases, each graded against an expected-behavior rubric by an LLM judge, from Corsac's Foundation Model & API eval coverage.
How is the Fim Completions Beta eval scored?+
The judge rubric: Grade against expected.ideal_behavior and expected.rubric. Per-criterion pass requires mean >= 4.0 and no criterion below 3.
How many test cases does this eval pack include?+
The Fim Completions Beta pack for DeepSeek DeepSeek API contains 9 test cases. 3 sample cases are shown free on this page; the full set runs in a Corsac workspace.
How do I run this eval?+
Sign up for Corsac, connect your model or agent endpoint, and run the Fim Completions Beta pack as-is or after customizing thresholds. Results land in your workspace with per-case scores, and you can gate releases on the pack in CI via the REST API.
Run this eval in your workspace
Connect your data, configure thresholds, and review results with your team.