diff --git a/studio/backend/core/inference/external_provider.py b/studio/backend/core/inference/external_provider.py
index bdb56c33ab..c71ef9ee1c 100644
--- a/studio/backend/core/inference/external_provider.py
+++ b/studio/backend/core/inference/external_provider.py
@@ -904,6 +904,34 @@ class ExternalProviderClient:
         tools: Optional[list[dict[str, Any]]] = None,
         tool_choice: Optional[Any] = None,
         fast_mode: Optional[bool] = None,
+        typical_p: Optional[float] = None,
+        top_n_sigma: Optional[float] = None,
+        repeat_last_n: Optional[int] = None,
+        dynatemp_range: Optional[float] = None,
+        dynatemp_exponent: Optional[float] = None,
+        mirostat: Optional[int] = None,
+        mirostat_tau: Optional[float] = None,
+        mirostat_eta: Optional[float] = None,
+        top_a: Optional[float] = None,
+        dry_multiplier: Optional[float] = None,
+        dry_base: Optional[float] = None,
+        dry_allowed_length: Optional[int] = None,
+        dry_penalty_last_n: Optional[int] = None,
+        xtc_probability: Optional[float] = None,
+        xtc_threshold: Optional[float] = None,
+        min_keep: Optional[int] = None,
+        ignore_eos: Optional[bool] = None,
+        min_tokens: Optional[int] = None,
+        skip_special_tokens: Optional[bool] = None,
+        spaces_between_special_tokens: Optional[bool] = None,
+        include_stop_str_in_output: Optional[bool] = None,
+        truncate_prompt_tokens: Optional[int] = None,
+        n_keep: Optional[int] = None,
+        n_probs: Optional[int] = None,
+        cache_prompt: Optional[bool] = None,
+        return_tokens: Optional[bool] = None,
+        timings_per_token: Optional[bool] = None,
+        post_sampling_probs: Optional[bool] = None,
         stream: bool = True,
     ) -> AsyncGenerator[str, None]:
         """
@@ -1067,10 +1095,76 @@ class ExternalProviderClient:
         if parallel_tool_calls is not None:
             body["parallel_tool_calls"] = parallel_tool_calls
 
+        # Extended OAI-compat samplers (OpenRouter `top_a`, vLLM output
+        # knobs, llama.cpp samplers on custom proxies). Each is gated `is
+        # not None` so explicit 0/False reach the wire; `body_omit` below
+        # strips fields the upstream rejects.
+        if typical_p is not None:
+            body["typical_p"] = typical_p
+        if top_n_sigma is not None:
+            body["top_n_sigma"] = top_n_sigma
+        if repeat_last_n is not None:
+            body["repeat_last_n"] = repeat_last_n
+        if dynatemp_range is not None:
+            body["dynatemp_range"] = dynatemp_range
+        if dynatemp_exponent is not None:
+            body["dynatemp_exponent"] = dynatemp_exponent
+        if mirostat is not None:
+            body["mirostat"] = mirostat
+        if mirostat_tau is not None:
+            body["mirostat_tau"] = mirostat_tau
+        if mirostat_eta is not None:
+            body["mirostat_eta"] = mirostat_eta
+        if top_a is not None:
+            body["top_a"] = top_a
+        if dry_multiplier is not None:
+            body["dry_multiplier"] = dry_multiplier
+        if dry_base is not None:
+            body["dry_base"] = dry_base
+        if dry_allowed_length is not None:
+            body["dry_allowed_length"] = dry_allowed_length
+        if dry_penalty_last_n is not None:
+            body["dry_penalty_last_n"] = dry_penalty_last_n
+        if xtc_probability is not None:
+            body["xtc_probability"] = xtc_probability
+        if xtc_threshold is not None:
+            body["xtc_threshold"] = xtc_threshold
+        if min_keep is not None:
+            body["min_keep"] = min_keep
+        if ignore_eos is not None:
+            body["ignore_eos"] = ignore_eos
+        if min_tokens is not None:
+            body["min_tokens"] = min_tokens
+        if skip_special_tokens is not None:
+            body["skip_special_tokens"] = skip_special_tokens
+        if spaces_between_special_tokens is not None:
+            body["spaces_between_special_tokens"] = spaces_between_special_tokens
+        if include_stop_str_in_output is not None:
+            body["include_stop_str_in_output"] = include_stop_str_in_output
+        if truncate_prompt_tokens is not None:
+            body["truncate_prompt_tokens"] = truncate_prompt_tokens
+        if n_keep is not None:
+            body["n_keep"] = n_keep
+        if n_probs is not None:
+            body["n_probs"] = n_probs
+        if cache_prompt is not None:
+            body["cache_prompt"] = cache_prompt
+        if return_tokens is not None:
+            body["return_tokens"] = return_tokens
+        if timings_per_token is not None:
+            body["timings_per_token"] = timings_per_token
+        if post_sampling_probs is not None:
+            body["post_sampling_probs"] = post_sampling_probs
+
         # Drop body fields the provider's registry entry locks down
         # (e.g. Kimi k2.5/k2.6 only accept temperature=1, top_p=1).
+        # Also pop the renamed seed field so `body_omit=("seed",)` on a
+        # provider with `seed_field` rename still strips correctly.
+        _seed_field = provider_info.get("seed_field", "seed")
         for field in provider_info.get("body_omit", ()):
             body.pop(field, None)
+            if field == "seed" and _seed_field != "seed":
+                body.pop(_seed_field, None)
 
         # Kimi thinking is a top-level body field. kimi-k2-thinking is
         # always on (ignore the toggle); kimi-k2.6 defaults on, can be
diff --git a/studio/backend/core/inference/providers.py b/studio/backend/core/inference/providers.py
index ed0047d8da..46ad4cee4d 100644
--- a/studio/backend/core/inference/providers.py
+++ b/studio/backend/core/inference/providers.py
@@ -192,6 +192,9 @@ PROVIDER_REGISTRY: dict[str, dict[str, Any]] = {
         # Mistral renames OpenAI's `seed` to `random_seed` on
         # /v1/chat/completions. https://docs.mistral.ai/api/endpoint/chat
         "seed_field": "random_seed",
+        # Mistral's docs publish no max but third-party shims cap at 4;
+        # match OpenAI Chat's cap to avoid silent upstream truncation.
+        "stop_max": 4,
     },
     "kimi": {
         "display_name": "Kimi",
@@ -216,12 +219,13 @@ PROVIDER_REGISTRY: dict[str, dict[str, Any]] = {
         "notes": "Moonshot API key. China: use base URL https://api.moonshot.cn/v1",
         "model_id_allowlist": re.compile(r"^kimi-k2\.[56]$"),
         # k2.5/k2.6 are reasoning-class: API locks temperature=1, top_p=1,
-        # frequency_penalty; seed and parallel_tool_calls are undocumented
-        # and 400.
+        # frequency_penalty; presence_penalty / seed / parallel_tool_calls
+        # are undocumented in the Kimi chat schema.
         "body_omit": (
             "temperature",
             "top_p",
             "frequency_penalty",
+            "presence_penalty",
             "seed",
             "parallel_tool_calls",
         ),
diff --git a/studio/backend/routes/chat_history.py b/studio/backend/routes/chat_history.py
index 800873afba..4e4faeb1ba 100644
--- a/studio/backend/routes/chat_history.py
+++ b/studio/backend/routes/chat_history.py
@@ -99,6 +99,9 @@ class ChatExportResponse(BaseModel):
 
 
 class ChatInferenceSettings(BaseModel):
+    # extra="forbid" requires every persisted key to be listed. Keep
+    # aligned with PERSISTED_INFERENCE_PARAM_KEYS in
+    # studio/frontend/src/features/chat/stores/chat-runtime-store.ts.
     model_config = ConfigDict(extra = "forbid")
 
     temperature: Optional[float] = None
@@ -107,10 +110,6 @@ class ChatInferenceSettings(BaseModel):
     minP: Optional[float] = None
     repetitionPenalty: Optional[float] = None
     presencePenalty: Optional[float] = None
-    # New per-provider sampling knobs. extra="forbid" requires these
-    # to be listed; otherwise every save from the new frontend 422s.
-    # Keep aligned with InferenceParams in
-    # studio/frontend/src/features/chat/types/runtime.ts.
     frequencyPenalty: Optional[float] = Field(default = None, ge = -2.0, le = 2.0)
     seed: Optional[int] = None
     stop: Optional[list[str]] = None
@@ -122,6 +121,36 @@ class ChatInferenceSettings(BaseModel):
     maxTokens: Optional[float] = None
     systemPrompt: Optional[str] = None
     trustRemoteCode: Optional[bool] = None
+    fastMode: Optional[bool] = None
+    # Extended llama.cpp / vLLM / OpenRouter samplers exposed by PR #5711.
+    typicalP: Optional[float] = None
+    topNSigma: Optional[float] = None
+    repeatLastN: Optional[int] = None
+    dynatempRange: Optional[float] = None
+    dynatempExponent: Optional[float] = None
+    mirostat: Optional[int] = None
+    mirostatTau: Optional[float] = None
+    mirostatEta: Optional[float] = None
+    topA: Optional[float] = None
+    dryMultiplier: Optional[float] = None
+    dryBase: Optional[float] = None
+    dryAllowedLength: Optional[int] = None
+    dryPenaltyLastN: Optional[int] = None
+    xtcProbability: Optional[float] = None
+    xtcThreshold: Optional[float] = None
+    minKeep: Optional[int] = None
+    ignoreEos: Optional[bool] = None
+    minTokens: Optional[int] = None
+    skipSpecialTokens: Optional[bool] = None
+    spacesBetweenSpecialTokens: Optional[bool] = None
+    includeStopStrInOutput: Optional[bool] = None
+    truncatePromptTokens: Optional[int] = None
+    nKeep: Optional[int] = None
+    nProbs: Optional[int] = None
+    cachePrompt: Optional[bool] = None
+    returnTokens: Optional[bool] = None
+    timingsPerToken: Optional[bool] = None
+    postSamplingProbs: Optional[bool] = None
 
 
 class ChatPreset(BaseModel):
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index cc0e236b49..481c9239d7 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -2168,6 +2168,34 @@ async def _proxy_to_external_provider(
             tools = payload.tools,
             tool_choice = payload.tool_choice,
             fast_mode = payload.fast_mode,
+            typical_p = payload.typical_p,
+            top_n_sigma = payload.top_n_sigma,
+            repeat_last_n = payload.repeat_last_n,
+            dynatemp_range = payload.dynatemp_range,
+            dynatemp_exponent = payload.dynatemp_exponent,
+            mirostat = payload.mirostat,
+            mirostat_tau = payload.mirostat_tau,
+            mirostat_eta = payload.mirostat_eta,
+            top_a = payload.top_a,
+            dry_multiplier = payload.dry_multiplier,
+            dry_base = payload.dry_base,
+            dry_allowed_length = payload.dry_allowed_length,
+            dry_penalty_last_n = payload.dry_penalty_last_n,
+            xtc_probability = payload.xtc_probability,
+            xtc_threshold = payload.xtc_threshold,
+            min_keep = payload.min_keep,
+            ignore_eos = payload.ignore_eos,
+            min_tokens = payload.min_tokens,
+            skip_special_tokens = payload.skip_special_tokens,
+            spaces_between_special_tokens = payload.spaces_between_special_tokens,
+            include_stop_str_in_output = payload.include_stop_str_in_output,
+            truncate_prompt_tokens = payload.truncate_prompt_tokens,
+            n_keep = payload.n_keep,
+            n_probs = payload.n_probs,
+            cache_prompt = payload.cache_prompt,
+            return_tokens = payload.return_tokens,
+            timings_per_token = payload.timings_per_token,
+            post_sampling_probs = payload.post_sampling_probs,
             stream = payload.stream,
         )
         try:
@@ -6009,6 +6037,25 @@ def _build_openai_passthrough_body(payload, backend_ctx = None) -> dict:
         mirostat = payload.mirostat,
         mirostat_tau = payload.mirostat_tau,
         mirostat_eta = payload.mirostat_eta,
+        dry_multiplier = payload.dry_multiplier,
+        dry_base = payload.dry_base,
+        dry_allowed_length = payload.dry_allowed_length,
+        dry_penalty_last_n = payload.dry_penalty_last_n,
+        xtc_probability = payload.xtc_probability,
+        xtc_threshold = payload.xtc_threshold,
+        min_keep = payload.min_keep,
+        ignore_eos = payload.ignore_eos,
+        min_tokens = payload.min_tokens,
+        skip_special_tokens = payload.skip_special_tokens,
+        spaces_between_special_tokens = payload.spaces_between_special_tokens,
+        include_stop_str_in_output = payload.include_stop_str_in_output,
+        truncate_prompt_tokens = payload.truncate_prompt_tokens,
+        n_keep = payload.n_keep,
+        n_probs = payload.n_probs,
+        cache_prompt = payload.cache_prompt,
+        return_tokens = payload.return_tokens,
+        timings_per_token = payload.timings_per_token,
+        post_sampling_probs = payload.post_sampling_probs,
         tool_choice = tool_choice,
         response_format = _extract_response_format(payload),
         chat_template_kwargs = tpl_kwargs,
diff --git a/studio/backend/tests/test_sampling_params_routing.py b/studio/backend/tests/test_sampling_params_routing.py
index 72e7eab412..c3115bf065 100644
--- a/studio/backend/tests/test_sampling_params_routing.py
+++ b/studio/backend/tests/test_sampling_params_routing.py
@@ -239,22 +239,20 @@ def test_anthropic_rejects_openai_only_knobs(monkeypatch):
 def _drive_openai_compat(captured, **kwargs) -> dict:
     """Send through the default OAI-compat branch (NOT /v1/responses).
 
-    Use a non-OpenAI provider_type so the dispatcher takes the default
-    branch at the bottom of stream_chat_completion rather than the
-    Responses translator path that routes provider_type=="openai".
+    Use qwen so the dispatcher takes the default branch and the provider
+    inherits the default 16-stop cap (Mistral now caps at 4 per its
+    third-party shims).
     """
 
     async def run():
         client = ExternalProviderClient(
-            provider_type = "mistral",
-            base_url = "https://api.mistral.ai/v1",
+            provider_type = "qwen",
+            base_url = "https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
             api_key = "test-key",
         )
-        # mistral's OpenAI-compat /v1/chat/completions returns OpenAI
-        # SSE; a single DONE frame is enough to drain the stream.
         async for _ in client.stream_chat_completion(
             messages = [{"role": "user", "content": "hi"}],
-            model = "mistral-small-latest",
+            model = "qwen-plus",
             temperature = 0.5,
             top_p = 0.9,
             max_tokens = 64,
@@ -278,10 +276,36 @@ def test_openai_compat_forwards_frequency_penalty(monkeypatch):
 
 
 def test_openai_compat_forwards_seed(monkeypatch):
+    """qwen has no seed_field override so seed forwards as `seed`."""
     captured = _install_mock(monkeypatch, sse_payload = _oai_done_payload())
     body = _drive_openai_compat(captured, seed = 12345)
-    # Default OAI-compat provider (mistral here) renames seed to
-    # random_seed via provider registry's seed_field.
+    assert body.get("seed") == 12345, body
+
+
+def test_mistral_renames_seed_to_random_seed(monkeypatch):
+    """Mistral's registry sets seed_field="random_seed" so the OAI seed
+    is renamed on the wire. https://docs.mistral.ai/api/endpoint/chat"""
+    captured = _install_mock(monkeypatch, sse_payload = _oai_done_payload())
+
+    async def run():
+        client = ExternalProviderClient(
+            provider_type = "mistral",
+            base_url = "https://api.mistral.ai/v1",
+            api_key = "mistral-test",
+        )
+        async for _ in client.stream_chat_completion(
+            messages = [{"role": "user", "content": "hi"}],
+            model = "mistral-small-latest",
+            temperature = 0.5,
+            top_p = 0.9,
+            max_tokens = 64,
+            seed = 12345,
+        ):
+            pass
+        await client.close()
+
+    _drive(run())
+    body = captured["body"]
     assert body.get("random_seed") == 12345, body
     assert "seed" not in body, body
 
@@ -349,8 +373,8 @@ def test_openai_compat_forwards_stop_array(monkeypatch):
 
 
 def test_openai_compat_truncates_stop_to_default_cap(monkeypatch):
-    """Default OAI-compat cap is 16 (DeepSeek and Mistral both accept
-    that many); only OpenAI Chat has a tighter 4-entry hard limit."""
+    """Default OAI-compat cap is 16 (Qwen, DeepSeek, HuggingFace, custom);
+    OpenAI Chat / OpenRouter / Gemini / Mistral have tighter 4-entry caps."""
     captured = _install_mock(monkeypatch, sse_payload = _oai_done_payload())
     body = _drive_openai_compat(captured, stop = [f"s{i}" for i in range(20)])
     assert len(body.get("stop", [])) == 16, body
@@ -358,6 +382,33 @@ def test_openai_compat_truncates_stop_to_default_cap(monkeypatch):
     assert body["stop"][-1] == "s15"
 
 
+def test_mistral_stop_cap_is_4(monkeypatch):
+    """Mistral's docs publish no max but third-party shims cap at 4;
+    match OpenAI Chat's cap to avoid silent upstream truncation."""
+    captured = _install_mock(monkeypatch, sse_payload = _oai_done_payload())
+
+    async def run():
+        client = ExternalProviderClient(
+            provider_type = "mistral",
+            base_url = "https://api.mistral.ai/v1",
+            api_key = "mistral-test",
+        )
+        async for _ in client.stream_chat_completion(
+            messages = [{"role": "user", "content": "hi"}],
+            model = "mistral-small-latest",
+            temperature = 0.5,
+            top_p = 0.9,
+            max_tokens = 64,
+            stop = [f"S{i}" for i in range(8)],
+        ):
+            pass
+        await client.close()
+
+    _drive(run())
+    body = captured["body"]
+    assert body.get("stop") == ["S0", "S1", "S2", "S3"], body
+
+
 def test_openai_compat_stop_dedup_and_drop_empties(monkeypatch):
     """Duplicates and empties shouldn't eat into the cap."""
     captured = _install_mock(monkeypatch, sse_payload = _oai_done_payload())
@@ -590,9 +641,9 @@ def test_kimi_web_search_bypass_forwards_new_sampling_fields(monkeypatch):
     assert "top_p" not in body, body
     assert "seed" not in body, body
     assert "parallel_tool_calls" not in body, body
+    assert "presence_penalty" not in body, body
     # Knobs not on Kimi's drop-list forward through the bypass.
     assert body.get("stop") == ["END"], body
-    assert body.get("presence_penalty") == 0.5, body
 
 
 def test_kimi_web_search_uses_kimi_stop_cap_5(monkeypatch):
@@ -681,6 +732,72 @@ def test_gemini_stop_sequences_capped_to_5(monkeypatch):
     assert gen_config.get("stopSequences") == ["S0", "S1", "S2", "S3", "S4"], body
 
 
+def test_openrouter_forwards_top_a(monkeypatch):
+    """OpenRouter exposes `top_a` (the tail-cut sampler). The frontend
+    capability map lets the value through; verify the OAI-compat body
+    builder actually carries it to the wire."""
+    captured = _install_mock(monkeypatch, sse_payload = _oai_done_payload())
+
+    async def run():
+        client = ExternalProviderClient(
+            provider_type = "openrouter",
+            base_url = "https://openrouter.ai/api/v1",
+            api_key = "or-test",
+        )
+        async for _ in client.stream_chat_completion(
+            messages = [{"role": "user", "content": "hi"}],
+            model = "openai/gpt-4o",
+            temperature = 0.5,
+            top_p = 0.9,
+            max_tokens = 64,
+            top_a = 0.25,
+        ):
+            pass
+        await client.close()
+
+    _drive(run())
+    body = captured["body"]
+    assert body.get("top_a") == 0.25, body
+
+
+def test_vllm_forwards_output_shape_knobs(monkeypatch):
+    """vLLM accepts skip_special_tokens / spaces_between_special_tokens /
+    include_stop_str_in_output / truncate_prompt_tokens. Verify the
+    OAI-compat external proxy forwards them to the wire body."""
+    captured = _install_mock(monkeypatch, sse_payload = _oai_done_payload())
+
+    async def run():
+        client = ExternalProviderClient(
+            provider_type = "vllm",
+            base_url = "https://vllm.example.com/v1",
+            api_key = "vllm-test",
+        )
+        async for _ in client.stream_chat_completion(
+            messages = [{"role": "user", "content": "hi"}],
+            model = "Qwen/Qwen3-4B",
+            temperature = 0.5,
+            top_p = 0.9,
+            max_tokens = 64,
+            skip_special_tokens = False,
+            spaces_between_special_tokens = False,
+            include_stop_str_in_output = True,
+            truncate_prompt_tokens = 2048,
+            min_tokens = 8,
+            ignore_eos = True,
+        ):
+            pass
+        await client.close()
+
+    _drive(run())
+    body = captured["body"]
+    assert body.get("skip_special_tokens") is False, body
+    assert body.get("spaces_between_special_tokens") is False, body
+    assert body.get("include_stop_str_in_output") is True, body
+    assert body.get("truncate_prompt_tokens") == 2048, body
+    assert body.get("min_tokens") == 8, body
+    assert body.get("ignore_eos") is True, body
+
+
 def test_kimi_drops_stop_strings_over_32_bytes(monkeypatch):
     """Kimi limits each stop string to <= 32 bytes per
     https://platform.kimi.ai/docs/api/chat. Drop overlong entries
@@ -837,27 +954,52 @@ def test_responses_to_chat_bridge_omits_unset_parallel_tool_calls():
 def test_chat_settings_payload_accepts_new_sampling_keys():
     """ChatSettingsPayload has extra="forbid" so the new keys must be
     listed explicitly; otherwise every settings save with any of them
-    422s. Pin the round-trip."""
+    422s. Pin the round-trip for every persisted sampler the frontend
+    can emit (see PERSISTED_INFERENCE_PARAM_KEYS in chat-runtime-store.ts)."""
     from routes.chat_history import ChatSettingsPayload
 
-    parsed = ChatSettingsPayload.model_validate(
-        {
-            "inferenceParams": {
-                "frequencyPenalty": 0.7,
-                "seed": 42,
-                "stop": ["END"],
-                "serviceTier": "standard_only",
-                "parallelToolCalls": False,
-            }
-        }
-    )
+    inference = {
+        "frequencyPenalty": 0.7,
+        "seed": 42,
+        "stop": ["END"],
+        "serviceTier": "standard_only",
+        "parallelToolCalls": False,
+        "fastMode": True,
+        # Extended llama.cpp / vLLM / OpenRouter samplers.
+        "typicalP": 0.85,
+        "topNSigma": 2.5,
+        "repeatLastN": 64,
+        "dynatempRange": 0.3,
+        "dynatempExponent": 1.2,
+        "mirostat": 2,
+        "mirostatTau": 4.0,
+        "mirostatEta": 0.15,
+        "topA": 0.2,
+        "dryMultiplier": 0.8,
+        "dryBase": 1.75,
+        "dryAllowedLength": 2,
+        "dryPenaltyLastN": -1,
+        "xtcProbability": 0.5,
+        "xtcThreshold": 0.1,
+        "minKeep": 5,
+        "ignoreEos": True,
+        "minTokens": 10,
+        "skipSpecialTokens": False,
+        "spacesBetweenSpecialTokens": False,
+        "includeStopStrInOutput": True,
+        "truncatePromptTokens": 1024,
+        "nKeep": -1,
+        "nProbs": 5,
+        "cachePrompt": False,
+        "returnTokens": True,
+        "timingsPerToken": True,
+        "postSamplingProbs": True,
+    }
+    parsed = ChatSettingsPayload.model_validate({"inferenceParams": inference})
     ip = parsed.inferenceParams
     assert ip is not None
-    assert ip.frequencyPenalty == 0.7
-    assert ip.seed == 42
-    assert ip.stop == ["END"]
-    assert ip.serviceTier == "standard_only"
-    assert ip.parallelToolCalls is False
+    for key, expected in inference.items():
+        assert getattr(ip, key) == expected, f"{key} did not round-trip"
 
 
 # ── Local /v1/messages: disable_parallel_tool_use translation ──────────
@@ -1093,7 +1235,7 @@ def test_local_passthrough_forwards_dry_xtc_min_keep_eos_min_tokens():
     assert body.get("min_tokens") == 16
 
     # Unset = absent from body. Matches the upstream "use default"
-    # contract — llama-server / vLLM apply their own defaults instead.
+    # contract: llama-server / vLLM apply their own defaults instead.
     body2 = route_mod._build_passthrough_payload(
         openai_messages = [{"role": "user", "content": "hi"}],
         openai_tools = None,
diff --git a/studio/frontend/src/features/chat/api/chat-adapter.ts b/studio/frontend/src/features/chat/api/chat-adapter.ts
index 5ec6cf40a9..11ef3b4700 100644
--- a/studio/frontend/src/features/chat/api/chat-adapter.ts
+++ b/studio/frontend/src/features/chat/api/chat-adapter.ts
@@ -2335,7 +2335,12 @@ export function createOpenAIStreamAdapter(): ChatModelAdapter {
             ...(params.postSamplingProbs === true
               ? { post_sampling_probs: true }
               : {}),
-            parallel_tool_calls: params.parallelToolCalls,
+            // Forward only on explicit opt-out (default true on every
+            // backend; default omit keeps wire-shape stable for users
+            // who never opened the new settings panel).
+            ...(params.parallelToolCalls === false
+              ? { parallel_tool_calls: false }
+              : {}),
             image_base64: imageBase64,
             audio_base64: audioBase64,
             cancel_id: cancelId,
diff --git a/studio/frontend/src/features/chat/chat-settings-sheet.tsx b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
index 2abc423ae5..ab6d68e3b2 100644
--- a/studio/frontend/src/features/chat/chat-settings-sheet.tsx
+++ b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
@@ -133,21 +133,11 @@ export function InfoHint({ children }: { children: ReactNode }) {
   );
 }
 
-/**
- * Editable numeric value display.
- *
- * Renders as a single <input> that *looks* like text by default —
- * transparent background, no border, no ring — and only shows a faint
- * surface tint on hover/focus to signal editability. When unfocused,
- * the input shows the formatted display string (`displayValue ?? value`,
- * so labels like "Off" / "Max" still render); on focus, it switches to
- * the raw numeric value, selects it, and accepts free text input.
- * Commit happens on blur or Enter; Escape reverts. The clamp-to-range
- * happens on commit so users can type intermediate values without the
- * input fighting them mid-keystroke. Single component shared by every
- * slider value and the Context Length input so the click-to-edit
- * affordance is consistent across the panel.
- */
+/** Editable numeric value display: transparent text-like input that
+ *  shows formatted display on blur (so "Off"/"Max" labels render) and
+ *  switches to the raw number on focus. Commits on blur/Enter, reverts
+ *  on Escape, clamps on commit. Shared by every slider value + the
+ *  Context Length input. */
 function snapToStep(
   value: number,
   step: number,
@@ -1431,7 +1421,19 @@ export function ChatSettingsPanel({
                   </InfoHint>
                 </div>
                 <Select
-                  value={params.serviceTier ?? "auto"}
+                  value={
+                    // Fall back to "auto" when the persisted tier is not
+                    // legal for the active provider (e.g. "priority" saved
+                    // on OpenAI, then user switched to Anthropic which only
+                    // accepts auto|standard_only). Without this Radix Select
+                    // shows a blank trigger.
+                    params.serviceTier &&
+                    (serviceTierOptions as readonly ServiceTier[]).includes(
+                      params.serviceTier,
+                    )
+                      ? params.serviceTier
+                      : "auto"
+                  }
                   onValueChange={(value) => {
                     // Store "auto" verbatim: Anthropic distinguishes
                     // omitted (provider default) from auto (Priority Tier opt-in).
@@ -1484,7 +1486,7 @@ export function ChatSettingsPanel({
                 max={32768}
                 step={128}
                 onChange={set("maxSeqLength")}
-                info="Maximum context window size in tokens — input prompt plus generated output combined. Capped by the model's trained limit."
+                info="Maximum context window in tokens (prompt plus generated output). Capped by the model's trained limit."
               />
             )}
             <ParamSlider
diff --git a/studio/frontend/src/features/chat/provider-capabilities.ts b/studio/frontend/src/features/chat/provider-capabilities.ts
index 0707ee6895..58341850d9 100644
--- a/studio/frontend/src/features/chat/provider-capabilities.ts
+++ b/studio/frontend/src/features/chat/provider-capabilities.ts
@@ -553,7 +553,7 @@ const OPENAI_COMPAT_BASE: ProviderCapabilities = {
 // plus the permissive `custom` preset. Exposes the full llama.cpp
 // sampler chain (typical_p / top_n_sigma / mirostat / dynatemp /
 // repeat_last_n) per the upstream server README. Not used for vLLM or
-// Ollama — see VLLM_OLLAMA_CAPABILITIES below.
+// Ollama; see VLLM_OLLAMA_CAPABILITIES below.
 const LLAMA_CPP_CAPABILITIES: ProviderCapabilities = {
   temperature: true,
   topP: true,
@@ -630,8 +630,8 @@ const VLLM_CAPABILITIES: ProviderCapabilities = {
 };
 
 // Ollama OAI translator (openai/openai.go FromChatRequest) only copies
-// the documented OpenAI subset on /v1/chat/completions — top_k / min_p
-// / repetition_penalty / ignore_eos / min_tokens / the 4 vLLM output
+// the documented OpenAI subset on /v1/chat/completions: top_k, min_p,
+// repetition_penalty, ignore_eos, min_tokens, and the 4 vLLM output
 // knobs all silently drop on this path. (Native /api/chat would forward
 // them via `options`, but Studio uses /v1.)
 const OLLAMA_CAPABILITIES: ProviderCapabilities = {
@@ -974,7 +974,7 @@ const PROVIDER_CAPABILITIES: Record<string, ProviderCapabilities> = {
     postSamplingProbs: false,
   },
   // DeepSeek schema (api-docs.deepseek.com/api/create-chat-completion)
-  // lists temperature/top_p/stop only — no seed or parallel_tool_calls.
+  // lists temperature/top_p/stop only; no seed or parallel_tool_calls.
   // Presence/frequency are deprecated. Reasoner ids additionally ignore
   // temperature/top_p; getProviderCapabilities downshifts them.
   deepseek: {
@@ -1023,7 +1023,7 @@ const PROVIDER_CAPABILITIES: Record<string, ProviderCapabilities> = {
   openrouter: OPENROUTER_CAPABILITIES,
   // llama_cpp + custom: first-party llama-server, full chain.
   // vllm: OAI subset + top_k/min_p/repetition_penalty/seed.
-  // ollama: stricter — OAI translator drops top_k/min_p/rep_pen too.
+  // ollama: stricter (OAI translator drops top_k/min_p/rep_pen too).
   custom: LLAMA_CPP_CAPABILITIES,
   llama_cpp: LLAMA_CPP_CAPABILITIES,
   vllm: VLLM_CAPABILITIES,
@@ -1055,8 +1055,8 @@ export function getProviderCapabilities(
 
 const DEFAULT_EFFORT_LEVELS = ["low", "medium", "high"] as const;
 // OpenRouter ids with no non-reasoning mode. (google/gemini-pro-latest
-// was dropped — gateway 404s; don't re-pin to a versioned id that
-// may rotate again.)
+// was dropped: gateway 404s; don't re-pin to a versioned id that may
+// rotate again.)
 const OPENROUTER_MANDATORY_REASONING_MODELS = new Set([
   "baidu/cobuddy:free",
   "inclusionai/ring-2.6-1t:free",
@@ -1377,7 +1377,7 @@ export interface ExternalReasoningResolveOptions {
   baseUrl?: string | null;
 }
 
-// vLLM has no per-model reasoning signal on OpenAI-compat — pin via user toggle.
+// vLLM has no per-model reasoning signal on OpenAI-compat; pin via user toggle.
 function resolveConnectionLevelReasoning(
   normalizedProvider: string,
   options: ExternalReasoningResolveOptions | undefined,
diff --git a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
index ca2ee65029..ddb3d71aa4 100644
--- a/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
+++ b/studio/frontend/src/features/chat/stores/chat-runtime-store.ts
@@ -416,6 +416,35 @@ const PERSISTED_INFERENCE_PARAM_KEYS = [
   "systemPrompt",
   "trustRemoteCode",
   "fastMode",
+  // Extended llama.cpp / vLLM / OpenRouter samplers exposed by PR #5711.
+  "typicalP",
+  "topNSigma",
+  "repeatLastN",
+  "dynatempRange",
+  "dynatempExponent",
+  "mirostat",
+  "mirostatTau",
+  "mirostatEta",
+  "topA",
+  "dryMultiplier",
+  "dryBase",
+  "dryAllowedLength",
+  "dryPenaltyLastN",
+  "xtcProbability",
+  "xtcThreshold",
+  "minKeep",
+  "ignoreEos",
+  "minTokens",
+  "skipSpecialTokens",
+  "spacesBetweenSpecialTokens",
+  "includeStopStrInOutput",
+  "truncatePromptTokens",
+  "nKeep",
+  "nProbs",
+  "cachePrompt",
+  "returnTokens",
+  "timingsPerToken",
+  "postSamplingProbs",
 ] as const satisfies readonly PersistedInferenceParamKey[];
 
 const SCALAR_SETTING_KEYS = [
diff --git a/studio/frontend/src/features/chat/types/runtime.ts b/studio/frontend/src/features/chat/types/runtime.ts
index 0e88ac0e0c..056749c43e 100644
--- a/studio/frontend/src/features/chat/types/runtime.ts
+++ b/studio/frontend/src/features/chat/types/runtime.ts
@@ -52,11 +52,11 @@ export interface InferenceParams {
   dryAllowedLength: number | null;
   /** 0 disables, -1 = ctx-size. */
   dryPenaltyLastN: number | null;
-  /** llama.cpp XTC — probability is the master switch (0 disables). */
+  /** llama.cpp XTC: probability is the master switch (0 disables). */
   xtcProbability: number | null;
   /** Default 0.1. */
   xtcThreshold: number | null;
-  /** llama.cpp `min_keep` — min tokens past all filters. 0 disables. */
+  /** llama.cpp `min_keep`: min tokens past all filters. 0 disables. */
   minKeep: number | null;
   /** Continue past EOS. llama.cpp + vLLM. */
   ignoreEos: boolean | null;
@@ -72,7 +72,7 @@ export interface InferenceParams {
   truncatePromptTokens: number | null;
   /** llama.cpp `n_keep`. 0 disables, -1 = keep all. */
   nKeep: number | null;
-  /** llama.cpp `n_probs` — top-N token probabilities per token. */
+  /** llama.cpp `n_probs`: top-N token probabilities per token. */
   nProbs: number | null;
   /** llama.cpp `cache_prompt`. Default true; forward only when false. */
   cachePrompt: boolean | null;