feat(eval): let env_args.model override --model in vf-eval

docs/evaluation.md

@@ -127,6 +127,23 @@ env.set_concurrency(256)
 
 The `renderer` client type requires the optional renderer package. Install it with `uv add "verifiers[renderers]"` before running evals with `--api-client-type renderer`.
 
+#### Model precedence
+
+`--model` / `-m` sets the inference client's model. Custom environments that
+need to know that model can accept `model` in `load_environment()` / their
+constructor, or read the injected `model` environment kwarg, instead of
+requiring users to repeat the same value in `--env-args`.
+
+To use a different model inside the environment than the one driving inference,
+pass it explicitly via `--env-args`:
+
+```bash
+prime eval run my-env -m google/gemma-3-27b-it -a '{"model": "qwen/qwen3-14b"}'
+```
+
+That override changes only the environment's view of `model`; the inference
+client still uses `--model`.
+
 For convenience, define model endpoints in `./configs/endpoints.toml` to avoid repeating URL and key flags.
 
 ```toml

tests/scripts/test_eval_model_kwarg.py

@@ -0,0 +1,33 @@
+from verifiers.scripts.eval import build_eval_config
+
+
+def test_resolved_model_lands_in_extra_env_kwargs(monkeypatch):
+    raw = {
+        "env_id": "math-python",
+        "model": "openai/gpt-4.1-mini",
+        "api_base_url": "https://example.test/v1",
+        "api_key_var": "OPENAI_API_KEY",
+    }
+    monkeypatch.setenv("OPENAI_API_KEY", "sk-test")
+
+    cfg = build_eval_config(raw)
+
+    assert cfg.model == "openai/gpt-4.1-mini"
+    assert cfg.extra_env_kwargs.get("model") == "openai/gpt-4.1-mini"
+
+
+def test_env_args_model_overrides_for_env_but_not_client(monkeypatch):
+    raw = {
+        "env_id": "math-python",
+        "model": "openai/gpt-4.1-mini",
+        "env_args": {"model": "qwen/qwen3-14b"},
+        "api_base_url": "https://example.test/v1",
+        "api_key_var": "OPENAI_API_KEY",
+    }
+    monkeypatch.setenv("OPENAI_API_KEY", "sk-test")
+
+    cfg = build_eval_config(raw)
+
+    assert cfg.model == "openai/gpt-4.1-mini"
+    assert cfg.env_args.get("model") == "qwen/qwen3-14b"
+    assert cfg.extra_env_kwargs.get("model") is None