diff --git a/hermes_cli/model_normalize.py b/hermes_cli/model_normalize.py
index 31bd5ff3c5..21d10c5e6f 100644
--- a/hermes_cli/model_normalize.py
+++ b/hermes_cli/model_normalize.py
@@ -104,6 +104,7 @@ _MATCHING_PREFIX_STRIP_PROVIDERS: frozenset[str] = frozenset({
     "xiaomi",
     "arcee",
     "ollama-cloud",
+    "nebius-token-factory",
     "custom",
     "gemini",
     "xai",
diff --git a/hermes_cli/models.py b/hermes_cli/models.py
index b545ffea65..f838df6247 100644
--- a/hermes_cli/models.py
+++ b/hermes_cli/models.py
@@ -1423,6 +1423,11 @@ _PROVIDER_ALIASES = {
     "actual-computer": "actual",
     "actualcomputer": "actual",
     "aci": "actual",
+    "nebius": "nebius-token-factory",
+    "nebius-tokenfactory": "nebius-token-factory",
+    "nebius-tf": "nebius-token-factory",
+    "token-factory": "nebius-token-factory",
+    "tokenfactory": "nebius-token-factory",
     "minimax-china": "minimax-cn",
     "minimax_cn": "minimax-cn",
     "minimax-portal": "minimax-oauth",
diff --git a/hermes_cli/providers.py b/hermes_cli/providers.py
index 2a02909ff6..caf6947e43 100644
--- a/hermes_cli/providers.py
+++ b/hermes_cli/providers.py
@@ -224,6 +224,12 @@ HERMES_OVERLAYS: Dict[str, HermesOverlay] = {
         base_url_override="https://api.upstage.ai/v1",
         base_url_env_var="UPSTAGE_BASE_URL",
     ),
+    "nebius-token-factory": HermesOverlay(
+        transport="openai_chat",
+        extra_env_vars=("NEBIUS_API_KEY", "NEBIUS_TOKEN_FACTORY_API_KEY"),
+        base_url_override="https://api.tokenfactory.nebius.com/v1",
+        base_url_env_var="NEBIUS_BASE_URL",
+    ),
     "ollama-cloud": HermesOverlay(
         transport="openai_chat",
         base_url_override="https://ollama.com/v1",
@@ -404,6 +410,12 @@ ALIASES: Dict[str, str] = {
     "actual-computer": "actual",
     "actualcomputer": "actual",
     "aci": "actual",
+    # Nebius Token Factory
+    "nebius": "nebius-token-factory",
+    "nebius-tokenfactory": "nebius-token-factory",
+    "nebius-tf": "nebius-token-factory",
+    "token-factory": "nebius-token-factory",
+    "tokenfactory": "nebius-token-factory",
 
     # Local server aliases → virtual "local" concept (resolved via user config)
     "lmstudio": "lmstudio",
@@ -432,6 +444,7 @@ _LABEL_OVERRIDES: Dict[str, str] = {
     "upstage": "Upstage Solar",
     "actual": "Actual Computer",
     "tencent-tokenhub": "Tencent TokenHub",
+    "nebius-token-factory": "Nebius Token Factory",
     "lmstudio": "LM Studio",
     "local": "Local endpoint",
     "bedrock": "AWS Bedrock",
diff --git a/plugins/model-providers/nebius-token-factory/__init__.py b/plugins/model-providers/nebius-token-factory/__init__.py
new file mode 100644
index 0000000000..19567c55f8
--- /dev/null
+++ b/plugins/model-providers/nebius-token-factory/__init__.py
@@ -0,0 +1,104 @@
+"""Nebius Token Factory provider profile."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from providers import register_provider
+from providers.base import ProviderProfile
+
+
+def _flat_model_name(model: str | None) -> str:
+    """Return a lowercase model id, tolerating vendor-prefixed IDs."""
+    return (model or "").strip().rsplit("/", 1)[-1].lower()
+
+
+def _model_supports_reasoning_effort(model: str | None) -> bool:
+    """Conservative allowlist for Nebius models that expose reasoning effort."""
+    model_name = _flat_model_name(model)
+    if not model_name:
+        return False
+    return any(
+        marker in model_name
+        for marker in (
+            "deepseek-r1",
+            "deepseek-v4",
+            "deepseek-reasoner",
+            "gpt-oss",
+            "glm-5",
+            "kimi-k2",
+            "minimax-m2",
+            "qwen3",
+        )
+    )
+
+
+class NebiusTokenFactoryProfile(ProviderProfile):
+    """Nebius Token Factory - top-level reasoning_effort."""
+
+    def build_api_kwargs_extras(
+        self,
+        *,
+        reasoning_config: dict | None = None,
+        model: str | None = None,
+        supports_reasoning: bool = False,
+        **context: Any,
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        if not supports_reasoning and not _model_supports_reasoning_effort(model):
+            return {}, {}
+
+        if isinstance(reasoning_config, dict):
+            enabled = reasoning_config.get("enabled", True)
+            raw_effort = reasoning_config.get("effort", "medium")
+        else:
+            enabled = True
+            raw_effort = "medium"
+
+        effort = str(raw_effort or "medium").strip().lower()
+        if enabled is False or effort in {"none", "off", "disabled"}:
+            return {}, {}
+        if effort in {"xhigh", "max"}:
+            effort = "high"
+        elif effort == "minimal":
+            effort = "low"
+        elif effort not in {"low", "medium", "high"}:
+            effort = "medium"
+
+        return {}, {"reasoning_effort": effort}
+
+
+nebius_token_factory = NebiusTokenFactoryProfile(
+    name="nebius-token-factory",
+    aliases=(
+        "nebius",
+        "nebius-tokenfactory",
+        "nebius-tf",
+        "token-factory",
+        "tokenfactory",
+    ),
+    display_name="Nebius Token Factory",
+    description="Nebius Token Factory — OpenAI-compatible inference",
+    signup_url="https://tokenfactory.nebius.com/",
+    env_vars=(
+        "NEBIUS_API_KEY",
+        "NEBIUS_TOKEN_FACTORY_API_KEY",
+        "NEBIUS_BASE_URL",
+    ),
+    base_url="https://api.tokenfactory.nebius.com/v1",
+    models_url="https://api.tokenfactory.nebius.com/v1/models",
+    auth_type="api_key",
+    default_aux_model="nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B",
+    fallback_models=(
+        "Qwen/Qwen3.5-397B-A17B-fast",
+        "deepseek-ai/DeepSeek-V4-Pro",
+        "zai-org/GLM-5.1",
+        "moonshotai/Kimi-K2.5-fast",
+        "MiniMaxAI/MiniMax-M2.5-fast",
+        "deepseek-ai/DeepSeek-V3.2-fast",
+        "NousResearch/Hermes-4-70B",
+        "openai/gpt-oss-120b-fast",
+        "meta-llama/Llama-3.3-70B-Instruct",
+    ),
+)
+
+register_provider(nebius_token_factory)
diff --git a/plugins/model-providers/nebius-token-factory/plugin.yaml b/plugins/model-providers/nebius-token-factory/plugin.yaml
new file mode 100644
index 0000000000..b76acc5d55
--- /dev/null
+++ b/plugins/model-providers/nebius-token-factory/plugin.yaml
@@ -0,0 +1,5 @@
+name: nebius-token-factory-provider
+kind: model-provider
+version: 1.0.0
+description: Nebius Token Factory OpenAI-compatible inference
+author: Nous Research
diff --git a/tests/hermes_cli/test_nebius_token_factory_provider.py b/tests/hermes_cli/test_nebius_token_factory_provider.py
new file mode 100644
index 0000000000..99983cfaeb
--- /dev/null
+++ b/tests/hermes_cli/test_nebius_token_factory_provider.py
@@ -0,0 +1,230 @@
+"""Focused tests for Nebius Token Factory provider wiring."""
+
+from __future__ import annotations
+
+from hermes_cli.auth import (
+    PROVIDER_REGISTRY,
+    resolve_api_key_provider_credentials,
+    resolve_provider,
+)
+from hermes_cli.model_normalize import normalize_model_for_provider
+from hermes_cli.models import (
+    CANONICAL_PROVIDERS,
+    _PROVIDER_ALIASES,
+    _PROVIDER_LABELS,
+    normalize_provider,
+    provider_model_ids,
+)
+
+
+def test_nebius_provider_profile_loads():
+    from providers import get_provider_profile
+
+    profile = get_provider_profile("nebius-token-factory")
+    assert profile is not None
+    assert profile.name == "nebius-token-factory"
+    assert profile.display_name == "Nebius Token Factory"
+    assert profile.base_url == "https://api.tokenfactory.nebius.com/v1"
+    assert profile.models_url == "https://api.tokenfactory.nebius.com/v1/models"
+    assert profile.env_vars[:2] == (
+        "NEBIUS_API_KEY",
+        "NEBIUS_TOKEN_FACTORY_API_KEY",
+    )
+    assert profile.default_aux_model == "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B"
+    assert "Qwen/Qwen3.5-397B-A17B-fast" in profile.fallback_models
+
+
+def test_nebius_aliases_resolve(monkeypatch):
+    monkeypatch.setenv("NEBIUS_API_KEY", "nebius-test-key")
+
+    for alias in (
+        "nebius",
+        "nebius-tokenfactory",
+        "nebius-tf",
+        "token-factory",
+        "tokenfactory",
+    ):
+        assert resolve_provider(alias) == "nebius-token-factory"
+        assert normalize_provider(alias) == "nebius-token-factory"
+        assert _PROVIDER_ALIASES[alias] == "nebius-token-factory"
+
+
+def test_nebius_provider_registry_and_credentials(monkeypatch):
+    monkeypatch.setenv("NEBIUS_API_KEY", "nebius-secret")
+    monkeypatch.setenv("NEBIUS_BASE_URL", "https://custom.nebius.example/v1")
+
+    pconfig = PROVIDER_REGISTRY["nebius-token-factory"]
+    assert pconfig.id == "nebius-token-factory"
+    assert pconfig.name == "Nebius Token Factory"
+    assert pconfig.auth_type == "api_key"
+    assert pconfig.inference_base_url == "https://api.tokenfactory.nebius.com/v1"
+    assert pconfig.api_key_env_vars == (
+        "NEBIUS_API_KEY",
+        "NEBIUS_TOKEN_FACTORY_API_KEY",
+    )
+    assert pconfig.base_url_env_var == "NEBIUS_BASE_URL"
+
+    creds = resolve_api_key_provider_credentials("nebius-token-factory")
+    assert creds["provider"] == "nebius-token-factory"
+    assert creds["api_key"] == "nebius-secret"
+    assert creds["base_url"] == "https://custom.nebius.example/v1"
+
+
+def test_nebius_canonical_provider_and_label():
+    slugs = [p.slug for p in CANONICAL_PROVIDERS]
+    assert "nebius-token-factory" in slugs
+    assert _PROVIDER_LABELS["nebius-token-factory"] == "Nebius Token Factory"
+
+
+def test_nebius_provider_module_overlay():
+    from hermes_cli.providers import (
+        HERMES_OVERLAYS,
+        determine_api_mode,
+        get_label,
+        get_provider,
+        normalize_provider as normalize_provider_in_providers,
+    )
+
+    overlay = HERMES_OVERLAYS["nebius-token-factory"]
+    assert overlay.transport == "openai_chat"
+    assert overlay.base_url_override == "https://api.tokenfactory.nebius.com/v1"
+    assert overlay.base_url_env_var == "NEBIUS_BASE_URL"
+
+    provider = get_provider("nebius")
+    assert provider is not None
+    assert provider.id == "nebius-token-factory"
+    assert provider.api_key_env_vars == (
+        "NEBIUS_API_KEY",
+        "NEBIUS_TOKEN_FACTORY_API_KEY",
+    )
+    assert provider.base_url == "https://api.tokenfactory.nebius.com/v1"
+    assert normalize_provider_in_providers("token-factory") == "nebius-token-factory"
+    assert get_label("nebius-token-factory") == "Nebius Token Factory"
+    assert determine_api_mode(
+        "nebius-token-factory",
+        "https://api.tokenfactory.nebius.com/v1",
+    ) == "chat_completions"
+
+
+def test_nebius_model_catalog_prefers_live_profile_fetch(monkeypatch):
+    from providers import get_provider_profile
+
+    profile = get_provider_profile("nebius-token-factory")
+    assert profile is not None
+    monkeypatch.setattr(
+        "hermes_cli.auth.resolve_api_key_provider_credentials",
+        lambda provider_id: {
+            "provider": provider_id,
+            "api_key": "nebius-live-key",
+            "base_url": "https://api.tokenfactory.nebius.com/v1",
+            "source": "NEBIUS_API_KEY",
+        },
+    )
+    monkeypatch.setattr(
+        profile,
+        "fetch_models",
+        lambda *, api_key=None, timeout=8.0: [
+            "deepseek-ai/DeepSeek-V4-Pro",
+            "NousResearch/Hermes-4-70B",
+        ],
+    )
+
+    assert provider_model_ids("nebius-token-factory") == [
+        "deepseek-ai/DeepSeek-V4-Pro",
+        "NousResearch/Hermes-4-70B",
+    ]
+
+
+def test_nebius_model_catalog_falls_back_to_profile_models(monkeypatch):
+    from providers import get_provider_profile
+
+    profile = get_provider_profile("nebius-token-factory")
+    assert profile is not None
+    monkeypatch.setattr(
+        "hermes_cli.auth.resolve_api_key_provider_credentials",
+        lambda provider_id: {
+            "provider": provider_id,
+            "api_key": "nebius-live-key",
+            "base_url": "https://api.tokenfactory.nebius.com/v1",
+            "source": "NEBIUS_API_KEY",
+        },
+    )
+    monkeypatch.setattr(profile, "fetch_models", lambda *, api_key=None, timeout=8.0: None)
+
+    assert provider_model_ids("nebius") == list(profile.fallback_models)
+
+
+def test_nebius_reasoning_models_emit_top_level_reasoning_effort():
+    from providers import get_provider_profile
+
+    profile = get_provider_profile("nebius-token-factory")
+    assert profile is not None
+
+    extra_body, top_level = profile.build_api_kwargs_extras(
+        reasoning_config={"enabled": True, "effort": "xhigh"},
+        model="openai/gpt-oss-120b-fast",
+    )
+    assert extra_body == {}
+    assert top_level == {"reasoning_effort": "high"}
+
+
+def test_nebius_reasoning_defaults_to_medium_for_known_reasoning_model():
+    from providers import get_provider_profile
+
+    profile = get_provider_profile("nebius-token-factory")
+    assert profile is not None
+
+    extra_body, top_level = profile.build_api_kwargs_extras(
+        reasoning_config=None,
+        model="deepseek-ai/DeepSeek-V4-Pro",
+    )
+    assert extra_body == {}
+    assert top_level == {"reasoning_effort": "medium"}
+
+
+def test_nebius_reasoning_skips_disabled_and_non_reasoning_models():
+    from providers import get_provider_profile
+
+    profile = get_provider_profile("nebius-token-factory")
+    assert profile is not None
+
+    assert profile.build_api_kwargs_extras(
+        reasoning_config={"enabled": False, "effort": "high"},
+        model="deepseek-ai/DeepSeek-V4-Pro",
+    ) == ({}, {})
+    assert profile.build_api_kwargs_extras(
+        reasoning_config={"enabled": True, "effort": "high"},
+        model="meta-llama/Llama-3.3-70B-Instruct",
+    ) == ({}, {})
+
+
+def test_nebius_transport_emits_top_level_reasoning_effort():
+    from agent.transports.chat_completions import ChatCompletionsTransport
+    from providers import get_provider_profile
+
+    profile = get_provider_profile("nebius-token-factory")
+    assert profile is not None
+
+    kwargs = ChatCompletionsTransport().build_kwargs(
+        model="deepseek-ai/DeepSeek-V4-Pro",
+        messages=[{"role": "user", "content": "ping"}],
+        tools=None,
+        provider_profile=profile,
+        reasoning_config={"enabled": True, "effort": "low"},
+        base_url="https://api.tokenfactory.nebius.com/v1",
+        provider_name="nebius-token-factory",
+    )
+    assert kwargs["reasoning_effort"] == "low"
+    assert "extra_body" not in kwargs
+
+
+def test_nebius_model_normalization_strips_canonical_and_alias_prefixes():
+    model = "Qwen/Qwen3.5-397B-A17B-fast"
+    assert normalize_model_for_provider(
+        f"nebius-token-factory/{model}", "nebius-token-factory"
+    ) == model
+    assert normalize_model_for_provider(f"nebius/{model}", "nebius-token-factory") == model
+    assert normalize_model_for_provider(f"nebius/{model}", "nebius") == model
+    assert normalize_model_for_provider(
+        "openai/gpt-oss-120b-fast", "nebius-token-factory"
+    ) == "openai/gpt-oss-120b-fast"
diff --git a/tests/providers/test_plugin_discovery.py b/tests/providers/test_plugin_discovery.py
index 2c828631bd..b725764512 100644
--- a/tests/providers/test_plugin_discovery.py
+++ b/tests/providers/test_plugin_discovery.py
@@ -70,6 +70,7 @@ def test_all_profiles_register():
     for required in (
         "openrouter", "anthropic", "custom", "bedrock", "openai-codex",
         "minimax-oauth", "gmi", "xiaomi", "alibaba-coding-plan", "fireworks",
+        "nebius-token-factory",
     ):
         assert required in names, f"Missing profile: {required}"