faresobeid · Oct 4, 2026
diff --git a/README.md b/README.md
@@ -57,7 +57,6 @@ The trainer runs the PrimeRL model implementations under `src/prime_rl/trainer/m
 | Nemotron H (`nemotron_h`) | `nvidia/Nemotron-3-Nano-30B-A3B`, `nvidia/Nemotron-3-Super-120B-A12B`, … | yes | ✅ | ✅ |
 | Trinity (`afmoe`) | `arcee-ai/Trinity-Mini`, … | yes | ✅ | ✅ |
 | GLM-4 · GLM-4.5 MoE · INTELLECT-3 (`glm4_moe`) | `THUDM/GLM-4-9B-0414`, `zai-org/GLM-4.5-Air`, `zai-org/GLM-4.5`, `PrimeIntellect/INTELLECT-3`, … | yes | ✅ | ✅ |
-| GPT-OSS (`gpt_oss`) | `unsloth/gpt-oss-20b-BF16`, … | yes | ✅ | ✅ |
 
 ## Setup
 

diff --git a/docs/advanced.md b/docs/advanced.md
@@ -29,12 +29,8 @@ The trainer only runs `prime-rl`'s own model implementations, selected from the
 | Nemotron H | `nvidia/Nemotron-3-Nano-30B-A3B`, … | ✅ | ❌ |
 | Trinity (AFMoE) | `arcee-ai/Trinity-Mini`, … | ✅ | ✅ |
 | GLM-4 / GLM-4.5 / INTELLECT-3 | `THUDM/GLM-4-9B-0414`, `zai-org/GLM-4.5`, `PrimeIntellect/INTELLECT-3`, … | ✅ | ✅ |
-| GPT-OSS | `unsloth/gpt-oss-20b-BF16`, … | ✅ | ✅ |
 | DeepSeek V4 | `deepseek-ai/DeepSeek-V4-Flash-0731` | ✅ | ✅ |
 
-GPT-OSS uses FlashAttention 4 with learned attention sinks. Training requires SM90 or SM100/SM110 GPUs
-and a BF16 checkpoint such as `unsloth/gpt-oss-20b-BF16`; the original MXFP4 checkpoints are not supported.
-
 ### Low-precision training
 
 Dense linear precision and routed-expert precision are configured independently. `[trainer.model.quantization]` applies only to dense `Linear` modules:
@@ -48,7 +44,7 @@ Dense linear precision and routed-expert precision are configured independently.
 - `type = "deepgemm_fp8"` (requires DeepGEMM and SM90+)
 - `type = "mxfp8"` (requires `prime-kernels`, torchao, and SM100)
 
-SonicMoE uses the upstream `sonic-moe` package (`uv sync --extra sonic-moe`) for fused BF16 expert computation. The supported model is Qwen3 MoE with the `gate_up` model fusion enabled. Backend selection requires fused gate/up weights, standard SwiGLU, and bias-free experts; incompatible expert structures raise an error during setup. It uses the same router and local, torch EP, or DeepEP dispatch as other compute backends:
+SonicMoE uses the upstream `sonic-moe` package (`uv sync --extra sonic-moe`) for fused BF16 expert computation. The supported model is Qwen3 MoE with the `gate_up` model fusion enabled. Backend selection requires fused gate/up weights and standard SwiGLU; incompatible expert structures raise an error during setup. It uses the same router and local, torch EP, or DeepEP dispatch as other compute backends:
 
 ```toml
 [trainer.model]

diff --git a/packages/prime-rl-configs/src/prime_rl/utils/parsers.py b/packages/prime-rl-configs/src/prime_rl/utils/parsers.py
@@ -12,7 +12,6 @@
     (re.compile(r"^zai-org/GLM-4\.7"), "glm47"),
     (re.compile(r"^zai-org/GLM-5"), "glm47"),
     (re.compile(r"^MiniMaxAI/MiniMax-M2"), "minimax_m2"),
-    (re.compile(r"^openai/gpt-oss"), "openai"),
     (re.compile(r"^poolside/Laguna"), "poolside_v1"),
     (re.compile(r"^PrimeIntellect/INTELLECT-3"), "qwen3_coder"),
     (re.compile(r"^nvidia/NVIDIA-Nemotron-3"), "qwen3_coder"),
@@ -32,7 +31,6 @@
     (re.compile(r"^deepseek-ai/DeepSeek-V3\.[12]"), "deepseek_r1"),
     (re.compile(r"^zai-org/GLM-"), "glm45"),
     (re.compile(r"^MiniMaxAI/MiniMax-M2"), "minimax_m2_append_think"),
-    # openai/gpt-oss has no entry here — vLLM's harmony serving path splits reasoning natively.
     (re.compile(r"^poolside/Laguna"), "poolside_v1"),
     (re.compile(r"^PrimeIntellect/INTELLECT-3"), "deepseek_r1"),
     (re.compile(r"^nvidia/NVIDIA-Nemotron-3-Super"), "nemotron_v3"),

diff --git a/src/prime_rl/inference/patches.py b/src/prime_rl/inference/patches.py
@@ -9,10 +9,8 @@ def apply_shared_vllm_patches():
     vLLM swallows plugin load failures (``load_plugins_by_group`` logs and continues), so a broken
     entry-point target silently skips ALL of these patches.
     """
-    from prime_rl.inference.vllm.gpt_oss_weight_loading import patch_gpt_oss_weight_loading
     from prime_rl.inference.vllm.qwen38_weight_loading import patch_qwen38_weight_loading
 
-    patch_gpt_oss_weight_loading()
     patch_qwen38_weight_loading()
     _patch_lora_key_prefix()
     _patch_qwen35_moe_lora_format()

diff --git a/src/prime_rl/inference/vllm/gpt_oss_weight_loading.py b/src/prime_rl/inference/vllm/gpt_oss_weight_loading.py
diff --git a/src/prime_rl/trainer/lora.py b/src/prime_rl/trainer/lora.py
@@ -14,7 +14,6 @@
     set_multilora_scaling,
 )
 from prime_rl.trainer.models.layers.lora.multi_moe import (
-    MultiLoRAGptOssGroupedExperts,
     MultiLoRAGroupedExperts,
     MultiLoRANonGatedGroupedExperts,
 )
@@ -224,9 +223,6 @@ def apply_lora_to_model(model: nn.Module, config: LoRAConfig) -> None:
     lora_state = setup_lora_state(config, torch.device("cuda", get_world().local_rank))
     if isinstance(model, PreTrainedModelPrimeRL):
         lora_state.register_adapter_state_dict_converter(type(model).convert_adapter_to_hf)
-    uses_gpt_oss_moe_adapter = (
-        isinstance(model, PreTrainedModelPrimeRL) and getattr(model.config, "model_type", None) == "gpt_oss"
-    )
 
     from torch.distributed.fsdp import FSDPModule
 
@@ -259,9 +255,7 @@ def apply_lora_to_model(model: nn.Module, config: LoRAConfig) -> None:
             )
         # Handle GroupedExperts (MoE)
         elif isinstance(base_module, GroupedExperts):
-            if uses_gpt_oss_moe_adapter:
-                wrapper = MultiLoRAGptOssGroupedExperts
-            elif base_module.gate_proj is not None:
+            if base_module.gate_proj is not None:
                 wrapper = MultiLoRAGroupedExperts
             else:
                 wrapper = MultiLoRANonGatedGroupedExperts

diff --git a/src/prime_rl/trainer/models/__init__.py b/src/prime_rl/trainer/models/__init__.py
@@ -14,7 +14,6 @@
 from prime_rl.trainer.models.deepseek_v4 import DeepseekV4Config, DeepseekV4ForCausalLM
 from prime_rl.trainer.models.glm4_moe import Glm4MoeConfig, Glm4MoeForCausalLM
 from prime_rl.trainer.models.glm_moe_dsa import GlmMoeDsaConfig, GlmMoeDsaForCausalLM
-from prime_rl.trainer.models.gpt_oss import GptOssConfig, GptOssForCausalLM
 from prime_rl.trainer.models.laguna import LagunaConfig, LagunaForCausalLM
 from prime_rl.trainer.models.layers.lm_head import PrimeLmOutput, cast_float_and_contiguous
 from prime_rl.trainer.models.llama import LlamaForCausalLM
@@ -41,7 +40,6 @@
 AutoConfig.register("deepseek_v4", DeepseekV4Config, exist_ok=True)
 AutoConfig.register("glm4_moe", Glm4MoeConfig, exist_ok=True)
 AutoConfig.register("glm_moe_dsa", GlmMoeDsaConfig, exist_ok=True)
-AutoConfig.register("gpt_oss", GptOssConfig, exist_ok=True)
 AutoConfig.register("laguna", LagunaConfig, exist_ok=True)
 AutoConfig.register("minimax_m2", MiniMaxM2Config, exist_ok=True)
 AutoConfig.register("nemotron_h", NemotronHConfig, exist_ok=True)
@@ -75,7 +73,6 @@
     (Qwen3_5MoeConfig, Qwen3_5ForCausalLM),
     (Qwen3_8FlashNextTextConfig, Qwen3_8FlashNextForCausalLM),
     (Qwen3_8FlashNextConfig, Qwen3_8FlashNextForCausalLM),
-    (GptOssConfig, GptOssForCausalLM),
 )
 for config_cls, model_cls in _CUSTOM_CAUSAL_LM_MODELS:
     _CUSTOM_CAUSAL_LM_MAPPING.register(config_cls, model_cls, exist_ok=True)

diff --git a/src/prime_rl/trainer/models/gpt_oss/__init__.py b/src/prime_rl/trainer/models/gpt_oss/__init__.py