Viewing enriched Twitter post
llama.cpp with MTP support makes local models fast enough to use as daily drivers 🚀 Qwen3.6-27B dense generation (on A10G): From 25 tok/s → 45 tok/s (+78%). Two flags on llama-server: --spec-type draft-mtp --spec-draft-n-max 2 https://t.co/hhslKpLE71
{
"media": [
{
"url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2056456757786869793/media_0.mp4",
"media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2056456757786869793/media_0.mp4",
"type": "video",
"filename": "media_0.mp4"
}
],
"processed_at": "2026-05-18T22:19:50.901805",
"pipeline_version": "2.0"
}{
"type": "tweet",
"id": "2056456757786869793",
"url": "https://x.com/victormustar/status/2056456757786869793",
"twitterUrl": "https://twitter.com/victormustar/status/2056456757786869793",
"text": "llama.cpp with MTP support makes local models fast enough to use as daily drivers 🚀\n\nQwen3.6-27B dense generation (on A10G): \nFrom 25 tok/s → 45 tok/s (+78%).\n\nTwo flags on llama-server:\n--spec-type draft-mtp --spec-draft-n-max 2 https://t.co/hhslKpLE71",
"source": "Twitter for iPhone",
"retweetCount": 19,
"replyCount": 9,
"likeCount": 125,
"quoteCount": 0,
"viewCount": 12234,
"createdAt": "Mon May 18 19:27:52 +0000 2026",
"lang": "en",
"bookmarkCount": 84,
"isReply": false,
"inReplyToId": null,
"conversationId": "2056456757786869793",
"displayTextRange": [
0,
229
],
"inReplyToUserId": null,
"inReplyToUsername": null,
"author": {
"type": "user",
"userName": "victormustar",
"url": "https://x.com/victormustar",
"twitterUrl": "https://twitter.com/victormustar",
"id": "784830007",
"name": "Victor M",
"isVerified": false,
"isBlueVerified": true,
"verifiedType": null,
"profilePicture": "https://pbs.twimg.com/profile_images/1099983311101984768/p7dZK4S__normal.jpg",
"coverPicture": "https://pbs.twimg.com/profile_banners/784830007/1677100565",
"description": "",
"location": "Paris, France",
"followers": 25429,
"following": 2211,
"status": "",
"canDm": true,
"canMediaTag": false,
"createdAt": "Mon Aug 27 14:52:18 +0000 2012",
"entities": {
"description": {
"urls": []
},
"url": {}
},
"fastFollowersCount": 0,
"favouritesCount": 27900,
"hasCustomTimelines": true,
"isTranslator": false,
"mediaCount": 1508,
"statusesCount": 7699,
"withheldInCountries": [],
"affiliatesHighlightedLabel": {},
"possiblySensitive": false,
"pinnedTweetIds": [],
"profile_bio": {
"description": "🤗 Head of Product @huggingface",
"entities": {
"description": {
"user_mentions": [
{
"id_str": "",
"indices": [
18,
30
],
"name": "",
"screen_name": "huggingface"
}
]
},
"url": {
"urls": [
{
"display_url": "hf.co/victor",
"expanded_url": "https://hf.co/victor",
"indices": [
0,
23
],
"url": "https://t.co/fgr8qRmdM5"
}
]
}
}
},
"isAutomated": false,
"automatedBy": null
},
"extendedEntities": {
"media": [
{
"additional_media_info": {
"monetizable": true
},
"display_url": "pic.twitter.com/hhslKpLE71",
"expanded_url": "https://twitter.com/victormustar/status/2056456757786869793/video/1",
"ext_media_availability": {
"status": "Available"
},
"id_str": "2056454141581328384",
"indices": [
230,
253
],
"media_key": "13_2056454141581328384",
"media_results": {
"id": "QXBpTWVkaWFSZXN1bHRzOgwABAoAARyJ/iVt1gAAAAA=",
"result": {
"__typename": "ApiMedia",
"id": "QXBpTWVkaWE6DAAECgABHIn+JW3WAAAAAA==",
"media_key": "13_2056454141581328384"
}
},
"media_url_https": "https://pbs.twimg.com/amplify_video_thumb/2056454141581328384/img/GVoU-ugvR0Bb2Omv.jpg",
"original_info": {
"focus_rects": [],
"height": 1080,
"width": 1380
},
"sizes": {
"large": {
"h": 1080,
"w": 1380
}
},
"type": "video",
"url": "https://t.co/hhslKpLE71",
"video_info": {
"aspect_ratio": [
23,
18
],
"duration_millis": 38050,
"variants": [
{
"content_type": "application/x-mpegURL",
"url": "https://video.twimg.com/amplify_video/2056454141581328384/pl/5nSZJZF55z6EVduF.m3u8?tag=27"
},
{
"bitrate": 256000,
"content_type": "video/mp4",
"url": "https://video.twimg.com/amplify_video/2056454141581328384/vid/avc1/344x270/tmqAyIxar1e9V5_e.mp4?tag=27"
},
{
"bitrate": 832000,
"content_type": "video/mp4",
"url": "https://video.twimg.com/amplify_video/2056454141581328384/vid/avc1/460x360/fDbHoKcdI-fnAJKf.mp4?tag=27"
},
{
"bitrate": 2176000,
"content_type": "video/mp4",
"url": "https://video.twimg.com/amplify_video/2056454141581328384/vid/avc1/920x720/oWGjUTLtLAL8Y8Kw.mp4?tag=27"
},
{
"bitrate": 10368000,
"content_type": "video/mp4",
"url": "https://video.twimg.com/amplify_video/2056454141581328384/vid/avc1/1380x1080/oS2nnIXGHgFe_1WU.mp4?tag=27"
}
]
}
}
]
},
"card": null,
"place": {},
"entities": {
"hashtags": [],
"symbols": [],
"timestamps": [],
"urls": [],
"user_mentions": []
},
"quoted_tweet": {
"type": "tweet",
"id": "2056391115469689330",
"url": "https://x.com/ggerganov/status/2056391115469689330",
"twitterUrl": "https://twitter.com/ggerganov/status/2056391115469689330",
"text": "llama.cpp adds MTP for the Qwen3.6 family\n\nThis is a significant milestone for the local AI ecosystem. The performance jump with these changes is massive and elevates local inference on commodity hardware further.\n\nSpecial thanks to Aman Gupta for leading this development!\n\nhttps://t.co/vjaMwEpIaR",
"source": "Twitter for iPhone",
"retweetCount": 89,
"replyCount": 31,
"likeCount": 629,
"quoteCount": 18,
"viewCount": 59460,
"createdAt": "Mon May 18 15:07:02 +0000 2026",
"lang": "en",
"bookmarkCount": 206,
"isReply": false,
"inReplyToId": null,
"conversationId": "2056391115469689330",
"displayTextRange": [
0,
273
],
"inReplyToUserId": null,
"inReplyToUsername": null,
"author": {
"type": "user",
"userName": "ggerganov",
"url": "https://x.com/ggerganov",
"twitterUrl": "https://twitter.com/ggerganov",
"id": "3300401027",
"name": "Georgi Gerganov",
"isVerified": false,
"isBlueVerified": true,
"verifiedType": null,
"profilePicture": "https://pbs.twimg.com/profile_images/1654097134315098113/zCZD0wYz_normal.jpg",
"coverPicture": "https://pbs.twimg.com/profile_banners/3300401027/1716922617",
"description": "",
"location": "",
"followers": 60264,
"following": 292,
"status": "",
"canDm": true,
"canMediaTag": false,
"createdAt": "Wed May 27 12:56:54 +0000 2015",
"entities": {
"description": {
"urls": []
},
"url": {}
},
"fastFollowersCount": 0,
"favouritesCount": 3182,
"hasCustomTimelines": true,
"isTranslator": false,
"mediaCount": 307,
"statusesCount": 1721,
"withheldInCountries": [],
"affiliatesHighlightedLabel": {},
"possiblySensitive": false,
"pinnedTweetIds": [
"2038632534414680223"
],
"profile_bio": {
"description": "24th at the Electrica puzzle challenge | building https://t.co/baTQS2bdia | engineer @huggingface",
"entities": {
"description": {
"urls": [
{
"display_url": "github.com/ggml-org",
"expanded_url": "https://github.com/ggml-org",
"indices": [
50,
73
],
"url": "https://t.co/baTQS2bdia"
}
],
"user_mentions": [
{
"id_str": "",
"indices": [
85,
97
],
"name": "",
"screen_name": "huggingface"
}
]
},
"url": {
"urls": [
{
"display_url": "github.com/ggerganov",
"expanded_url": "https://github.com/ggerganov",
"indices": [
0,
23
],
"url": "https://t.co/0H5L2twAKi"
}
]
}
}
},
"isAutomated": false,
"automatedBy": null
},
"extendedEntities": {},
"card": {
"binding_values": [
{
"key": "photo_image_full_size_large",
"value": {
"image_value": {
"height": 419,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=800x419",
"width": 800
}
}
},
{
"key": "thumbnail_image",
"value": {
"image_value": {
"height": 200,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=400x400",
"width": 400
}
}
},
{
"key": "description",
"value": {
"string_value": "Overview This PR adds support for MTP (Multi Token Prediction) heads. I tested this on Qwen3.6 27B and Qwen3.6 35BA3B but in principle it should work for any MTP model. I've posted the det..."
}
},
{
"key": "domain",
"value": {
"string_value": "github.com"
}
},
{
"key": "thumbnail_image_large",
"value": {
"image_value": {
"height": 300,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=600x600",
"width": 600
}
}
},
{
"key": "summary_photo_image_small",
"value": {
"image_value": {
"height": 202,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=386x202",
"width": 386
}
}
},
{
"key": "thumbnail_image_original",
"value": {
"image_value": {
"height": 600,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=orig",
"width": 1200
}
}
},
{
"key": "site",
"value": {
"scribe_key": "publisher_id",
"user_value": {
"id_str": "13334762",
"path": []
}
}
},
{
"key": "photo_image_full_size_small",
"value": {
"image_value": {
"height": 202,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=386x202",
"width": 386
}
}
},
{
"key": "summary_photo_image_large",
"value": {
"image_value": {
"height": 419,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=800x419",
"width": 800
}
}
},
{
"key": "thumbnail_image_small",
"value": {
"image_value": {
"height": 72,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=144x144",
"width": 144
}
}
},
{
"key": "thumbnail_image_x_large",
"value": {
"image_value": {
"height": 600,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=png&name=2048x2048_2_exp",
"width": 1200
}
}
},
{
"key": "photo_image_full_size_original",
"value": {
"image_value": {
"height": 600,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=orig",
"width": 1200
}
}
},
{
"key": "photo_image_full_size_alt_text",
"value": {
"string_value": "Overview This PR adds support for MTP (Multi Token Prediction) heads. I tested this on Qwen3.6 27B and Qwen3.6 35BA3B but in principle it should work for any MTP model. I've posted the detaile..."
}
},
{
"key": "vanity_url",
"value": {
"scribe_key": "vanity_url",
"string_value": "github.com"
}
},
{
"key": "photo_image_full_size",
"value": {
"image_value": {
"height": 314,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=600x314",
"width": 600
}
}
},
{
"key": "summary_photo_image_alt_text",
"value": {
"string_value": "Overview This PR adds support for MTP (Multi Token Prediction) heads. I tested this on Qwen3.6 27B and Qwen3.6 35BA3B but in principle it should work for any MTP model. I've posted the detaile..."
}
},
{
"key": "thumbnail_image_color",
"value": {
"image_color_value": {
"palette": [
{
"percentage": 91.67,
"rgb": {
"blue": 255,
"green": 255,
"red": 255
}
},
{
"percentage": 3.69,
"rgb": {
"blue": 118,
"green": 113,
"red": 108
}
},
{
"percentage": 1.81,
"rgb": {
"blue": 126,
"green": 75,
"red": 243
}
},
{
"percentage": 1,
"rgb": {
"blue": 179,
"green": 149,
"red": 248
}
},
{
"percentage": 0.34,
"rgb": {
"blue": 196,
"green": 234,
"red": 186
}
}
]
}
}
},
{
"key": "title",
"value": {
"string_value": "llama + spec: MTP Support by am17an · Pull Request #22673 · ggml-org/llama.cpp"
}
},
{
"key": "summary_photo_image_color",
"value": {
"image_color_value": {
"palette": [
{
"percentage": 91.67,
"rgb": {
"blue": 255,
"green": 255,
"red": 255
}
},
{
"percentage": 3.69,
"rgb": {
"blue": 118,
"green": 113,
"red": 108
}
},
{
"percentage": 1.81,
"rgb": {
"blue": 126,
"green": 75,
"red": 243
}
},
{
"percentage": 1,
"rgb": {
"blue": 179,
"green": 149,
"red": 248
}
},
{
"percentage": 0.34,
"rgb": {
"blue": 196,
"green": 234,
"red": 186
}
}
]
}
}
},
{
"key": "summary_photo_image_x_large",
"value": {
"image_value": {
"height": 600,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=png&name=2048x2048_2_exp",
"width": 1200
}
}
},
{
"key": "summary_photo_image",
"value": {
"image_value": {
"height": 314,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=600x314",
"width": 600
}
}
},
{
"key": "photo_image_full_size_color",
"value": {
"image_color_value": {
"palette": [
{
"percentage": 91.67,
"rgb": {
"blue": 255,
"green": 255,
"red": 255
}
},
{
"percentage": 3.69,
"rgb": {
"blue": 118,
"green": 113,
"red": 108
}
},
{
"percentage": 1.81,
"rgb": {
"blue": 126,
"green": 75,
"red": 243
}
},
{
"percentage": 1,
"rgb": {
"blue": 179,
"green": 149,
"red": 248
}
},
{
"percentage": 0.34,
"rgb": {
"blue": 196,
"green": 234,
"red": 186
}
}
]
}
}
},
{
"key": "photo_image_full_size_x_large",
"value": {
"image_value": {
"height": 600,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=png&name=2048x2048_2_exp",
"width": 1200
}
}
},
{
"key": "card_url",
"value": {
"scribe_key": "card_url",
"string_value": "https://t.co/vjaMwEpIaR"
}
},
{
"key": "summary_photo_image_original",
"value": {
"image_value": {
"height": 600,
"url": "https://pbs.twimg.com/card_img/2055828069081100288/YBRU1g7b?format=jpg&name=orig",
"width": 1200
}
}
}
],
"card_platform": {
"platform": {
"audience": {
"name": "production"
},
"device": {
"name": "iPhone",
"version": "13"
}
}
},
"name": "summary_large_image",
"url": "https://t.co/vjaMwEpIaR",
"user_refs_results": [
{
"rest_id": "13334762",
"result": {
"__typename": "User",
"action_counts": {
"favorites_count": 8693
},
"avatar": {
"image_url": "https://pbs.twimg.com/profile_images/2051404708766507008/ATxxTJXO_normal.jpg"
},
"banner": {
"image_url": "https://pbs.twimg.com/profile_banners/13334762/1765308302"
},
"core": {
"created_at": "Mon Feb 11 04:41:50 +0000 2008",
"name": "GitHub",
"screen_name": "github"
},
"dm_permissions": {
"can_dm": true
},
"exclusive_tweet_following": false,
"follow_request_sent": false,
"identity_profile_labels_highlighted_label": {},
"location": {
"location": "San Francisco, CA"
},
"media_permissions": {
"can_media_tag": true
},
"notifications_settings": {
"notifications_enabled": false
},
"pinned_items": {},
"privacy": {
"protected": false,
"suspended": false
},
"private_super_following": false,
"profile_bio": {
"description": "The AI-powered developer platform to build, scale, and deliver secure software.",
"entities": {
"description": {},
"url": {
"urls": [
{
"display_url": "github.com",
"expanded_url": "http://github.com",
"indices": [
0,
23
],
"url": "https://t.co/bbJgfyzKzp"
}
]
}
}
},
"profile_image_shape": "Square",
"profile_metadata": {
"profile_interstitial_type": "",
"profile_link_color": "981CEB"
},
"profile_translation": {
"translator_type_enum": "None"
},
"properties": {
"has_extended_profile": true,
"has_no_screen_name": false
},
"relationship_counts": {
"followers": 2646551,
"following": 334
},
"relationship_perspectives": {
"blocked_by": false,
"blocking": false,
"followed_by": false,
"following": false,
"live_following": false,
"muting": false
},
"rest_id": "13334762",
"smart_blocked_by": false,
"smart_blocking": false,
"super_follow_eligible": false,
"super_followed_by": false,
"super_following": false,
"tweet_counts": {
"media_tweets": 2974,
"tweets": 10545
},
"verification": {
"is_blue_verified": true,
"verified": false,
"verified_type": "Business"
},
"website": {
"url": "https://t.co/bbJgfyzKzp"
}
}
}
]
},
"place": {},
"entities": {
"hashtags": [],
"symbols": [],
"urls": [
{
"display_url": "github.com/ggml-org/llama…",
"expanded_url": "https://github.com/ggml-org/llama.cpp/pull/22673",
"indices": [
275,
298
],
"url": "https://t.co/vjaMwEpIaR"
}
],
"user_mentions": []
},
"quoted_tweet": null,
"retweeted_tweet": null,
"isLimitedReply": false,
"communityInfo": null,
"article": null
},
"retweeted_tweet": null,
"isLimitedReply": false,
"communityInfo": null,
"article": null
}