🐦 Twitter Post Details

Viewing enriched Twitter post

@dair_ai

Banger paper from NVIDIA. Agentic reasoning needs models that are not just capable, but efficient at long-context inference. The agent model layer is moving toward open, long-context, high-throughput architectures. This paper introduces Nemotron 3 Super, an open 120B parameter model with 12B active parameters, built as a hybrid Mamba-Attention Mixture-of-Experts architecture. The headline numbers are strong: up to 1M context length, comparable accuracy on common benchmarks, and up to 2.2x higher throughput than GPT-OSS-120B and 7.5x higher throughput than Qwen3.5-122B. The model combines several efficiency bets, including NVFP4 pretraining, LatentMoE for accuracy per FLOP and per parameter, and MTP layers for native speculative decoding. It is trained on 25 trillion tokens, then post-trained with supervised fine-tuning and RL. Paper: https://t.co/VcqUPjylzF Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c

Media 1
View on Twitter

📊 Media Metadata

{
  "media": [
    {
      "type": "photo",
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2044452957023047943/media_0.jpg",
      "filename": "media_0.jpg"
    }
  ],
  "processed_at": "2026-04-15T16:31:20.207910",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2044452957023047943",
  "url": "https://x.com/dair_ai/status/2044452957023047943",
  "twitterUrl": "https://twitter.com/dair_ai/status/2044452957023047943",
  "text": "Banger paper from NVIDIA.\n\nAgentic reasoning needs models that are not just capable, but efficient at long-context inference.\n\nThe agent model layer is moving toward open, long-context, high-throughput architectures.\n\nThis paper introduces Nemotron 3 Super, an open 120B parameter model with 12B active parameters, built as a hybrid Mamba-Attention Mixture-of-Experts architecture.\n\nThe headline numbers are strong: up to 1M context length, comparable accuracy on common benchmarks, and up to 2.2x higher throughput than GPT-OSS-120B and 7.5x higher throughput than Qwen3.5-122B.\n\nThe model combines several efficiency bets, including NVFP4 pretraining, LatentMoE for accuracy per FLOP and per parameter, and MTP layers for native speculative decoding. It is trained on 25 trillion tokens, then post-trained with supervised fine-tuning and RL.\n\nPaper: https://t.co/VcqUPjylzF\n\nLearn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c",
  "source": "Twitter for iPhone",
  "retweetCount": 1,
  "replyCount": 0,
  "likeCount": 4,
  "quoteCount": 0,
  "viewCount": 107,
  "createdAt": "Wed Apr 15 16:29:03 +0000 2026",
  "lang": "en",
  "bookmarkCount": 6,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2044452957023047943",
  "displayTextRange": [
    0,
    280
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "dair_ai",
    "url": "https://x.com/dair_ai",
    "twitterUrl": "https://twitter.com/dair_ai",
    "id": "889050642903293953",
    "name": "DAIR.AI",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/1643277398522187778/31dedbLo_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/889050642903293953/1773242460",
    "description": "",
    "location": "",
    "followers": 102856,
    "following": 1,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Sun Jul 23 09:12:45 +0000 2017",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 4342,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 193,
    "statusesCount": 3092,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2044066936045351317"
    ],
    "profile_bio": {
      "description": "Democratizing AI research, education, and technologies. New AI learning portal: https://t.co/LRnpZN7L4c",
      "entities": {
        "description": {
          "hashtags": [],
          "symbols": [],
          "urls": [
            {
              "display_url": "academy.dair.ai",
              "expanded_url": "https://academy.dair.ai/",
              "indices": [
                80,
                103
              ],
              "url": "https://t.co/LRnpZN7L4c"
            }
          ],
          "user_mentions": []
        },
        "url": {
          "urls": [
            {
              "display_url": "dair.ai",
              "expanded_url": "https://www.dair.ai/",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/lkqPZtMU5s"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "display_url": "pic.twitter.com/p116fMqkcy",
        "expanded_url": "https://twitter.com/dair_ai/status/2044452957023047943/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {
            "faces": []
          },
          "orig": {
            "faces": []
          }
        },
        "id_str": "2044452953973788678",
        "indices": [
          281,
          304
        ],
        "media_key": "3_2044452953973788678",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARxfWyExmxAGCgACHF9bIedbEQcAAA==",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAABCgABHF9bITGbEAYKAAIcX1sh51sRBwAA",
            "media_key": "3_2044452953973788678"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/HF9bITGbEAY0j5n.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 748,
              "w": 1336,
              "x": 0,
              "y": 0
            },
            {
              "h": 1336,
              "w": 1336,
              "x": 0,
              "y": 0
            },
            {
              "h": 1523,
              "w": 1336,
              "x": 0,
              "y": 0
            },
            {
              "h": 1792,
              "w": 896,
              "x": 0,
              "y": 0
            },
            {
              "h": 1792,
              "w": 1336,
              "x": 0,
              "y": 0
            }
          ],
          "height": 1792,
          "width": 1336
        },
        "sizes": {
          "large": {
            "h": 1792,
            "w": 1336
          }
        },
        "type": "photo",
        "url": "https://t.co/p116fMqkcy"
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "urls": [
      {
        "display_url": "arxiv.org/abs/2604.12374",
        "expanded_url": "https://arxiv.org/abs/2604.12374",
        "indices": [
          852,
          875
        ],
        "url": "https://t.co/VcqUPjylzF"
      },
      {
        "display_url": "academy.dair.ai",
        "expanded_url": "https://academy.dair.ai/",
        "indices": [
          928,
          951
        ],
        "url": "https://t.co/LRnpZN7L4c"
      }
    ],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "communityInfo": null,
  "article": null
}